mirror of https://mirror.osredm.com/root/redis.git
Async IO Threads (#13695)
## Introduction
Redis introduced IO Thread in 6.0, allowing IO threads to handle client
request reading, command parsing and reply writing, thereby improving
performance. The current IO thread implementation has a few drawbacks.
- The main thread is blocked during IO thread read/write operations and
must wait for all IO threads to complete their current tasks before it
can continue execution. In other words, the entire process is
synchronous. This prevents the efficient utilization of multi-core CPUs
for parallel processing.
- When the number of clients and requests increases moderately, it
causes all IO threads to reach full CPU utilization due to the busy wait
mechanism used by the IO threads. This makes it challenging for us to
determine which part of Redis has reached its bottleneck.
- When IO threads are enabled with TLS and io-threads-do-reads, a
disconnection of a connection with pending data may result in it being
assigned to multiple IO threads simultaneously. This can cause race
conditions and trigger assertion failures. Related issue:
redis#12540
Therefore, we designed an asynchronous IO threads solution. The IO
threads adopt an event-driven model, with the main thread dedicated to
command processing, meanwhile, the IO threads handle client read and
write operations in parallel.
## Implementation
### Overall
As before, we did not change the fact that all client commands must be
executed on the main thread, because Redis was originally designed to be
single-threaded, and processing commands in a multi-threaded manner
would inevitably introduce numerous race and synchronization issues. But
now each IO thread has independent event loop, therefore, IO threads can
use a multiplexing approach to handle client read and write operations,
eliminating the CPU overhead caused by busy-waiting.
the execution process can be briefly described as follows:
the main thread assigns clients to IO threads after accepting
connections, IO threads will notify the main thread when clients
finish reading and parsing queries, then the main thread processes
queries from IO threads and generates replies, IO threads handle
writing reply to clients after receiving clients list from main thread,
and then continue to handle client read and write events.
### Each IO thread has independent event loop
We now assign each IO thread its own event loop. This approach
eliminates the need for the main thread to perform the costly
`epoll_wait` operation for handling connections (except for specific
ones). Instead, the main thread processes requests from the IO threads
and hands them back once completed, fully offloading read and write
events to the IO threads.
Additionally, all TLS operations, including handling pending data, have
been moved entirely to the IO threads. This resolves the issue where
io-threads-do-reads could not be used with TLS.
### Event-notified client queue
To facilitate communication between the IO threads and the main thread,
we designed an event-notified client queue. Each IO thread and the main
thread have two such queues to store clients waiting to be processed.
These queues are also integrated with the event loop to enable handling.
We use pthread_mutex to ensure the safety of queue operations, as well
as data visibility and ordering, and race conditions are minimized, as
each IO thread and the main thread operate on independent queues,
avoiding thread suspension due to lock contention. And we implemented an
event notifier based on `eventfd` or `pipe` to support event-driven
handling.
### Thread safety
Since the main thread and IO threads can execute in parallel, we must
handle data race issues carefully.
**client->flags**
The primary tasks of IO threads are reading and writing, i.e.
`readQueryFromClient` and `writeToClient`. However, IO threads and the
main thread may concurrently modify or access `client->flags`, leading
to potential race conditions. To address this, we introduced an io-flags
variable to record operations performed by IO threads, thereby avoiding
race conditions on `client->flags`.
**Pause IO thread**
In the main thread, we may want to operate data of IO threads, maybe
uninstall event handler, access or operate query/output buffer or resize
event loop, we need a clean and safe context to do that. We pause IO
thread in `IOThreadBeforeSleep`, do some jobs and then resume it. To
avoid thread suspended, we use busy waiting to confirm the target
status. Besides we use atomic variable to make sure memory visibility
and ordering. We introduce these functions to pause/resume IO Threads as
below.
```
pauseIOThread, resumeIOThread
pauseAllIOThreads, resumeAllIOThreads
pauseIOThreadsRange, resumeIOThreadsRange
```
Testing has shown that `pauseIOThread` is highly efficient, allowing the
main thread to execute nearly 200,000 operations per second during
stress tests. Similarly, `pauseAllIOThreads` with 8 IO threads can
handle up to nearly 56,000 operations per second. But operations
performed between pausing and resuming IO threads must be quick;
otherwise, they could cause the IO threads to reach full CPU
utilization.
**freeClient and freeClientAsync**
The main thread may need to terminate a client currently running on an
IO thread, for example, due to ACL rule changes, reaching the output
buffer limit, or evicting a client. In such cases, we need to pause the
IO thread to safely operate on the client.
**maxclients and maxmemory-clients updating**
When adjusting `maxclients`, we need to resize the event loop for all IO
threads. Similarly, when modifying `maxmemory-clients`, we need to
traverse all clients to calculate their memory usage. To ensure safe
operations, we pause all IO threads during these adjustments.
**Client info reading**
The main thread may need to read a client’s fields to generate a
descriptive string, such as for the `CLIENT LIST` command or logging
purposes. In such cases, we need to pause the IO thread handling that
client. If information for all clients needs to be displayed, all IO
threads must be paused.
**Tracking redirect**
Redis supports the tracking feature and can even send invalidation
messages to a connection with a specified ID. But the target client may
be running on IO thread, directly manipulating the client’s output
buffer is not thread-safe, and the IO thread may not be aware that the
client requires a response. In such cases, we pause the IO thread
handling the client, modify the output buffer, and install a write event
handler to ensure proper handling.
**clientsCron**
In the `clientsCron` function, the main thread needs to traverse all
clients to perform operations such as timeout checks, verifying whether
they have reached the soft output buffer limit, resizing the
output/query buffer, or updating memory usage. To safely operate on a
client, the IO thread handling that client must be paused.
If we were to pause the IO thread for each client individually, the
efficiency would be very low. Conversely, pausing all IO threads
simultaneously would be costly, especially when there are many IO
threads, as clientsCron is invoked relatively frequently.
To address this, we adopted a batched approach for pausing IO threads.
At most, 8 IO threads are paused at a time. The operations mentioned
above are only performed on clients running in the paused IO threads,
significantly reducing overhead while maintaining safety.
### Observability
In the current design, the main thread always assigns clients to the IO
thread with the least clients. To clearly observe the number of clients
handled by each IO thread, we added the new section in INFO output. The
`INFO THREADS` section can show the client count for each IO thread.
```
# Threads
io_thread_0:clients=0
io_thread_1:clients=2
io_thread_2:clients=2
```
Additionally, in the `CLIENT LIST` output, we also added a field to
indicate the thread to which each client is assigned.
`id=244 addr=127.0.0.1:41870 laddr=127.0.0.1:6379 ... resp=2 lib-name=
lib-ver= io-thread=1`
## Trade-off
### Special Clients
For certain special types of clients, keeping them running on IO threads
would result in severe race issues that are difficult to resolve.
Therefore, we chose not to offload these clients to the IO threads.
For replica, monitor, subscribe, and tracking clients, main thread may
directly write them a reply when conditions are met. Race issues are
difficult to resolve, so we have them processed in the main thread. This
includes the Lua debug clients as well, since we may operate connection
directly.
For blocking client, after the IO thread reads and parses a command and
hands it over to the main thread, if the client is identified as a
blocking type, it will be remained in the main thread. Once the blocking
operation completes and the reply is generated, the client is
transferred back to the IO thread to send the reply and wait for event
triggers.
### Clients Eviction
To support client eviction, it is necessary to update each client’s
memory usage promptly during operations such as read, write, or command
execution. However, when a client operates on an IO thread, it is not
feasible to update the memory usage immediately due to the risk of data
races. As a result, memory usage can only be updated either in the main
thread while processing commands or in the `ClientsCron` periodically.
The downside of this approach is that updates might experience a delay
of up to one second, which could impact the precision of memory
management for eviction.
To avoid incorrectly evicting clients. We adopted a best-effort
compensation solution, when we decide to eviction a client, we update
its memory usage again before evicting, if the memory used by the client
does not decrease or memory usage bucket is not changed, then we will
evict it, otherwise, not evict it.
However, we have not completely solved this problem. Due to the delay in
memory usage updates, it may lead us to make incorrect decisions about
the need to evict clients.
### Defragment
In the majority of cases we do NOT use the data from argv directly in
the db.
1. key names
We store a copy that we allocate in the main thread, see `sdsdup()` in
`dbAdd()`.
2. hash key and value
We store key as hfield and store value as sds, see `hfieldNew()` and
`sdsdup()` in `hashTypeSet()`.
3. other datatypes
They don't even use SDS, so there is no reference issues.
But in some cases client the data from argv may be retain by the main
thread.
As a result, during fragmentation cleanup, we need to move allocations
from the IO thread’s arena to the main thread’s arena. We always
allocate new memory in the main thread’s arena, but the memory released
by IO threads may not yet have been reclaimed. This ultimately causes
the fragmentation rate to be higher compared to creating and allocating
entirely within a single thread.
The following cases below will lead to memory allocated by the IO thread
being kept by the main thread.
1. string related command: `append`, `getset`, `mset` and `set`.
If `tryObjectEncoding()` does not change argv, we will keep it directly
in the main thread, see the code in `tryObjectEncoding()`(specifically
`trimStringObjectIfNeeded()`)
2. block related command.
the key names will be kept in `c->db->blocking_keys`.
3. watch command
the key names will be kept in `c->db->watched_keys`.
4. [s]subscribe command
channel name will be kept in `serverPubSubChannels`.
5. script load command
script will be kept in `server.lua_scripts`.
7. some module API: `RM_RetainString`, `RM_HoldString`
Those issues will be handled in other PRs.
## Testing
### Functional Testing
The commit with enabling IO Threads has passed all TCL tests, but we did
some changes:
**Client query buffer**: In the original code, when using a reusable
query buffer, ownership of the query buffer would be released after the
command was processed. However, with IO threads enabled, the client
transitions from an IO thread to the main thread for processing. This
causes the ownership release to occur earlier than the command
execution. As a result, when IO threads are enabled, the client's
information will never indicate that a shared query buffer is in use.
Therefore, we skip the corresponding query buffer tests in this case.
**Defragment**: Add a new defragmentation test to verify the effect of
io threads on defragmentation.
**Command delay**: For deferred clients in TCL tests, due to clients
being assigned to different threads for execution, delays may occur. To
address this, we introduced conditional waiting: the process proceeds to
the next step only when the `client list` contains the corresponding
commands.
### Sanitizer Testing
The commit passed all TCL tests and reported no errors when compiled
with the `fsanitizer=thread` and `fsanitizer=address` options enabled.
But we made the following modifications: we suppressed the sanitizer
warnings for clients with watched keys when updating `client->flags`, we
think IO threads read `client->flags`, but never modify it or read the
`CLIENT_DIRTY_CAS` bit, main thread just only modifies this bit, so
there is no actual data race.
## Others
### IO thread number
In the new multi-threaded design, the main thread is primarily focused
on command processing to improve performance. Typically, the main thread
does not handle regular client I/O operations but is responsible for
clients such as replication and tracking clients. To avoid breaking
changes, we still consider the main thread as the first IO thread.
When the io-threads configuration is set to a low value (e.g., 2),
performance does not show a significant improvement compared to a
single-threaded setup for simple commands (such as SET or GET), as the
main thread does not consume much CPU for these simple operations. This
results in underutilized multi-core capacity. However, for more complex
commands, having a low number of IO threads may still be beneficial.
Therefore, it’s important to adjust the `io-threads` based on your own
performance tests.
Additionally, you can clearly monitor the CPU utilization of the main
thread and IO threads using `top -H -p $redis_pid`. This allows you to
easily identify where the bottleneck is. If the IO thread is the
bottleneck, increasing the `io-threads` will improve performance. If the
main thread is the bottleneck, the overall performance can only be
scaled by increasing the number of shards or replicas.
---------
Co-authored-by: debing.sun <debing.sun@redis.com>
Co-authored-by: oranagra <oran@redislabs.com>
This commit is contained in:
Yuan Wang
GitHub
parent
08c2b276fb
commit
64a40b20d9
31
redis.conf
31
redis.conf
|
|
@ -1291,38 +1291,27 @@ lazyfree-lazy-user-flush no
|
|||
# in different I/O threads. Since especially writing is so slow, normally
|
||||
# Redis users use pipelining in order to speed up the Redis performances per
|
||||
# core, and spawn multiple instances in order to scale more. Using I/O
|
||||
# threads it is possible to easily speedup two times Redis without resorting
|
||||
# threads it is possible to easily speedup several times Redis without resorting
|
||||
# to pipelining nor sharding of the instance.
|
||||
#
|
||||
# By default threading is disabled, we suggest enabling it only in machines
|
||||
# that have at least 4 or more cores, leaving at least one spare core.
|
||||
# Using more than 8 threads is unlikely to help much. We also recommend using
|
||||
# threaded I/O only if you actually have performance problems, with Redis
|
||||
# instances being able to use a quite big percentage of CPU time, otherwise
|
||||
# there is no point in using this feature.
|
||||
# We also recommend using threaded I/O only if you actually have performance
|
||||
# problems, with Redis instances being able to use a quite big percentage of
|
||||
# CPU time, otherwise there is no point in using this feature.
|
||||
#
|
||||
# So for instance if you have a four cores boxes, try to use 2 or 3 I/O
|
||||
# threads, if you have a 8 cores, try to use 6 threads. In order to
|
||||
# So for instance if you have a four cores boxes, try to use 3 I/O
|
||||
# threads, if you have a 8 cores, try to use 7 threads. In order to
|
||||
# enable I/O threads use the following configuration directive:
|
||||
#
|
||||
# io-threads 4
|
||||
#
|
||||
# Setting io-threads to 1 will just use the main thread as usual.
|
||||
# When I/O threads are enabled, we only use threads for writes, that is
|
||||
# to thread the write(2) syscall and transfer the client buffers to the
|
||||
# socket. However it is also possible to enable threading of reads and
|
||||
# protocol parsing using the following configuration directive, by setting
|
||||
# it to yes:
|
||||
# When I/O threads are enabled, we not only use threads for writes, that
|
||||
# is to thread the write(2) syscall and transfer the client buffers to the
|
||||
# socket, but also use threads for reads and protocol parsing.
|
||||
#
|
||||
# io-threads-do-reads no
|
||||
#
|
||||
# Usually threading reads doesn't help much.
|
||||
#
|
||||
# NOTE 1: This configuration directive cannot be changed at runtime via
|
||||
# CONFIG SET. Also, this feature currently does not work when SSL is
|
||||
# enabled.
|
||||
#
|
||||
# NOTE 2: If you want to test the Redis speedup using redis-benchmark, make
|
||||
# NOTE: If you want to test the Redis speedup using redis-benchmark, make
|
||||
# sure you also run the benchmark itself in threaded mode, using the
|
||||
# --threads option to match the number of Redis threads, otherwise you'll not
|
||||
# be able to notice the improvements.
|
||||
|
|
|
|||
|
|
@ -354,7 +354,7 @@ endif
|
|||
|
||||
REDIS_SERVER_NAME=redis-server$(PROG_SUFFIX)
|
||||
REDIS_SENTINEL_NAME=redis-sentinel$(PROG_SUFFIX)
|
||||
REDIS_SERVER_OBJ=threads_mngr.o adlist.o quicklist.o ae.o anet.o dict.o ebuckets.o mstr.o kvstore.o server.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o cluster_legacy.o crc16.o endianconv.o slowlog.o eval.o bio.o rio.o rand.o memtest.o syscheck.o crcspeed.o crccombine.o crc64.o bitops.o sentinel.o notify.o setproctitle.o blocked.o hyperloglog.o latency.o sparkline.o redis-check-rdb.o redis-check-aof.o geo.o lazyfree.o module.o evict.o expire.o geohash.o geohash_helper.o childinfo.o defrag.o siphash.o rax.o t_stream.o listpack.o localtime.o lolwut.o lolwut5.o lolwut6.o acl.o tracking.o socket.o tls.o sha256.o timeout.o setcpuaffinity.o monotonic.o mt19937-64.o resp_parser.o call_reply.o script_lua.o script.o functions.o function_lua.o commands.o strl.o connection.o unix.o logreqres.o
|
||||
REDIS_SERVER_OBJ=threads_mngr.o adlist.o quicklist.o ae.o anet.o dict.o ebuckets.o eventnotifier.o iothread.o mstr.o kvstore.o server.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o cluster_legacy.o crc16.o endianconv.o slowlog.o eval.o bio.o rio.o rand.o memtest.o syscheck.o crcspeed.o crccombine.o crc64.o bitops.o sentinel.o notify.o setproctitle.o blocked.o hyperloglog.o latency.o sparkline.o redis-check-rdb.o redis-check-aof.o geo.o lazyfree.o module.o evict.o expire.o geohash.o geohash_helper.o childinfo.o defrag.o siphash.o rax.o t_stream.o listpack.o localtime.o lolwut.o lolwut5.o lolwut6.o acl.o tracking.o socket.o tls.o sha256.o timeout.o setcpuaffinity.o monotonic.o mt19937-64.o resp_parser.o call_reply.o script_lua.o script.o functions.o function_lua.o commands.o strl.o connection.o unix.o logreqres.o
|
||||
REDIS_CLI_NAME=redis-cli$(PROG_SUFFIX)
|
||||
REDIS_CLI_OBJ=anet.o adlist.o dict.o redis-cli.o zmalloc.o release.o ae.o redisassert.o crcspeed.o crccombine.o crc64.o siphash.o crc16.o monotonic.o cli_common.o mt19937-64.o strl.o cli_commands.o
|
||||
REDIS_BENCHMARK_NAME=redis-benchmark$(PROG_SUFFIX)
|
||||
|
|
|
|||
41
src/ae.c
41
src/ae.c
|
|
@ -42,7 +42,7 @@
|
|||
#endif
|
||||
#endif
|
||||
|
||||
|
||||
#define INITIAL_EVENT 1024
|
||||
aeEventLoop *aeCreateEventLoop(int setsize) {
|
||||
aeEventLoop *eventLoop;
|
||||
int i;
|
||||
|
|
@ -50,8 +50,9 @@ aeEventLoop *aeCreateEventLoop(int setsize) {
|
|||
monotonicInit(); /* just in case the calling app didn't initialize */
|
||||
|
||||
if ((eventLoop = zmalloc(sizeof(*eventLoop))) == NULL) goto err;
|
||||
eventLoop->events = zmalloc(sizeof(aeFileEvent)*setsize);
|
||||
eventLoop->fired = zmalloc(sizeof(aeFiredEvent)*setsize);
|
||||
eventLoop->nevents = setsize < INITIAL_EVENT ? setsize : INITIAL_EVENT;
|
||||
eventLoop->events = zmalloc(sizeof(aeFileEvent)*eventLoop->nevents);
|
||||
eventLoop->fired = zmalloc(sizeof(aeFiredEvent)*eventLoop->nevents);
|
||||
if (eventLoop->events == NULL || eventLoop->fired == NULL) goto err;
|
||||
eventLoop->setsize = setsize;
|
||||
eventLoop->timeEventHead = NULL;
|
||||
|
|
@ -61,10 +62,11 @@ aeEventLoop *aeCreateEventLoop(int setsize) {
|
|||
eventLoop->beforesleep = NULL;
|
||||
eventLoop->aftersleep = NULL;
|
||||
eventLoop->flags = 0;
|
||||
memset(eventLoop->privdata, 0, sizeof(eventLoop->privdata));
|
||||
if (aeApiCreate(eventLoop) == -1) goto err;
|
||||
/* Events with mask == AE_NONE are not set. So let's initialize the
|
||||
* vector with it. */
|
||||
for (i = 0; i < setsize; i++)
|
||||
for (i = 0; i < eventLoop->nevents; i++)
|
||||
eventLoop->events[i].mask = AE_NONE;
|
||||
return eventLoop;
|
||||
|
||||
|
|
@ -102,20 +104,19 @@ void aeSetDontWait(aeEventLoop *eventLoop, int noWait) {
|
|||
*
|
||||
* Otherwise AE_OK is returned and the operation is successful. */
|
||||
int aeResizeSetSize(aeEventLoop *eventLoop, int setsize) {
|
||||
int i;
|
||||
|
||||
if (setsize == eventLoop->setsize) return AE_OK;
|
||||
if (eventLoop->maxfd >= setsize) return AE_ERR;
|
||||
if (aeApiResize(eventLoop,setsize) == -1) return AE_ERR;
|
||||
|
||||
eventLoop->events = zrealloc(eventLoop->events,sizeof(aeFileEvent)*setsize);
|
||||
eventLoop->fired = zrealloc(eventLoop->fired,sizeof(aeFiredEvent)*setsize);
|
||||
eventLoop->setsize = setsize;
|
||||
|
||||
/* Make sure that if we created new slots, they are initialized with
|
||||
* an AE_NONE mask. */
|
||||
for (i = eventLoop->maxfd+1; i < setsize; i++)
|
||||
eventLoop->events[i].mask = AE_NONE;
|
||||
/* If the current allocated space is larger than the requested size,
|
||||
* we need to shrink it to the requested size. */
|
||||
if (setsize < eventLoop->nevents) {
|
||||
eventLoop->events = zrealloc(eventLoop->events,sizeof(aeFileEvent)*setsize);
|
||||
eventLoop->fired = zrealloc(eventLoop->fired,sizeof(aeFiredEvent)*setsize);
|
||||
eventLoop->nevents = setsize;
|
||||
}
|
||||
return AE_OK;
|
||||
}
|
||||
|
||||
|
|
@ -147,6 +148,22 @@ int aeCreateFileEvent(aeEventLoop *eventLoop, int fd, int mask,
|
|||
errno = ERANGE;
|
||||
return AE_ERR;
|
||||
}
|
||||
|
||||
/* Resize the events and fired arrays if the file
|
||||
* descriptor exceeds the current number of events. */
|
||||
if (unlikely(fd >= eventLoop->nevents)) {
|
||||
int newnevents = eventLoop->nevents;
|
||||
newnevents = (newnevents * 2 > fd + 1) ? newnevents * 2 : fd + 1;
|
||||
newnevents = (newnevents > eventLoop->setsize) ? eventLoop->setsize : newnevents;
|
||||
eventLoop->events = zrealloc(eventLoop->events, sizeof(aeFileEvent) * newnevents);
|
||||
eventLoop->fired = zrealloc(eventLoop->fired, sizeof(aeFiredEvent) * newnevents);
|
||||
|
||||
/* Initialize new slots with an AE_NONE mask */
|
||||
for (int i = eventLoop->nevents; i < newnevents; i++)
|
||||
eventLoop->events[i].mask = AE_NONE;
|
||||
eventLoop->nevents = newnevents;
|
||||
}
|
||||
|
||||
aeFileEvent *fe = &eventLoop->events[fd];
|
||||
|
||||
if (aeApiAddEvent(eventLoop, fd, mask) == -1)
|
||||
|
|
|
|||
2
src/ae.h
2
src/ae.h
|
|
@ -79,6 +79,7 @@ typedef struct aeEventLoop {
|
|||
int maxfd; /* highest file descriptor currently registered */
|
||||
int setsize; /* max number of file descriptors tracked */
|
||||
long long timeEventNextId;
|
||||
int nevents; /* Size of Registered events */
|
||||
aeFileEvent *events; /* Registered events */
|
||||
aeFiredEvent *fired; /* Fired events */
|
||||
aeTimeEvent *timeEventHead;
|
||||
|
|
@ -87,6 +88,7 @@ typedef struct aeEventLoop {
|
|||
aeBeforeSleepProc *beforesleep;
|
||||
aeBeforeSleepProc *aftersleep;
|
||||
int flags;
|
||||
void *privdata[2];
|
||||
} aeEventLoop;
|
||||
|
||||
/* Prototypes */
|
||||
|
|
|
|||
|
|
@ -317,7 +317,7 @@ migrateCachedSocket* migrateGetSocket(client *c, robj *host, robj *port, long ti
|
|||
}
|
||||
|
||||
/* Create the connection */
|
||||
conn = connCreate(connTypeOfCluster());
|
||||
conn = connCreate(server.el, connTypeOfCluster());
|
||||
if (connBlockingConnect(conn, host->ptr, atoi(port->ptr), timeout)
|
||||
!= C_OK) {
|
||||
addReplyError(c,"-IOERR error or timeout connecting to the client");
|
||||
|
|
|
|||
|
|
@ -1262,7 +1262,7 @@ void clusterAcceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
|
|||
return;
|
||||
}
|
||||
|
||||
connection *conn = connCreateAccepted(connTypeOfCluster(), cfd, &require_auth);
|
||||
connection *conn = connCreateAccepted(server.el, connTypeOfCluster(), cfd, &require_auth);
|
||||
|
||||
/* Make sure connection is not in an error state */
|
||||
if (connGetState(conn) != CONN_STATE_ACCEPTING) {
|
||||
|
|
@ -4583,7 +4583,7 @@ static int clusterNodeCronHandleReconnect(clusterNode *node, mstime_t handshake_
|
|||
|
||||
if (node->link == NULL) {
|
||||
clusterLink *link = createClusterLink(node);
|
||||
link->conn = connCreate(connTypeOfCluster());
|
||||
link->conn = connCreate(server.el, connTypeOfCluster());
|
||||
connSetPrivateData(link->conn, link);
|
||||
if (connConnect(link->conn, node->ip, node->cport, server.bind_source_addr,
|
||||
clusterLinkConnectHandler) == C_ERR) {
|
||||
|
|
|
|||
|
|
@ -1239,6 +1239,9 @@ commandHistory CLIENT_LIST_History[] = {
|
|||
{"6.2.0","Added `argv-mem`, `tot-mem`, `laddr` and `redir` fields and the optional `ID` filter."},
|
||||
{"7.0.0","Added `resp`, `multi-mem`, `rbs` and `rbp` fields."},
|
||||
{"7.0.3","Added `ssub` field."},
|
||||
{"7.2.0","Added `lib-name` and `lib-ver` fields."},
|
||||
{"7.4.0","Added `watch` field."},
|
||||
{"8.0.0","Added `io-thread` field."},
|
||||
};
|
||||
#endif
|
||||
|
||||
|
|
@ -1546,7 +1549,7 @@ struct COMMAND_STRUCT CLIENT_Subcommands[] = {
|
|||
{MAKE_CMD("id","Returns the unique client ID of the connection.","O(1)","5.0.0",CMD_DOC_NONE,NULL,NULL,"connection",COMMAND_GROUP_CONNECTION,CLIENT_ID_History,0,CLIENT_ID_Tips,0,clientCommand,2,CMD_NOSCRIPT|CMD_LOADING|CMD_STALE|CMD_SENTINEL,ACL_CATEGORY_CONNECTION,CLIENT_ID_Keyspecs,0,NULL,0)},
|
||||
{MAKE_CMD("info","Returns information about the connection.","O(1)","6.2.0",CMD_DOC_NONE,NULL,NULL,"connection",COMMAND_GROUP_CONNECTION,CLIENT_INFO_History,0,CLIENT_INFO_Tips,1,clientCommand,2,CMD_NOSCRIPT|CMD_LOADING|CMD_STALE|CMD_SENTINEL,ACL_CATEGORY_CONNECTION,CLIENT_INFO_Keyspecs,0,NULL,0)},
|
||||
{MAKE_CMD("kill","Terminates open connections.","O(N) where N is the number of client connections","2.4.0",CMD_DOC_NONE,NULL,NULL,"connection",COMMAND_GROUP_CONNECTION,CLIENT_KILL_History,6,CLIENT_KILL_Tips,0,clientCommand,-3,CMD_ADMIN|CMD_NOSCRIPT|CMD_LOADING|CMD_STALE|CMD_SENTINEL,ACL_CATEGORY_CONNECTION,CLIENT_KILL_Keyspecs,0,NULL,1),.args=CLIENT_KILL_Args},
|
||||
{MAKE_CMD("list","Lists open connections.","O(N) where N is the number of client connections","2.4.0",CMD_DOC_NONE,NULL,NULL,"connection",COMMAND_GROUP_CONNECTION,CLIENT_LIST_History,6,CLIENT_LIST_Tips,1,clientCommand,-2,CMD_ADMIN|CMD_NOSCRIPT|CMD_LOADING|CMD_STALE|CMD_SENTINEL,ACL_CATEGORY_CONNECTION,CLIENT_LIST_Keyspecs,0,NULL,2),.args=CLIENT_LIST_Args},
|
||||
{MAKE_CMD("list","Lists open connections.","O(N) where N is the number of client connections","2.4.0",CMD_DOC_NONE,NULL,NULL,"connection",COMMAND_GROUP_CONNECTION,CLIENT_LIST_History,9,CLIENT_LIST_Tips,1,clientCommand,-2,CMD_ADMIN|CMD_NOSCRIPT|CMD_LOADING|CMD_STALE|CMD_SENTINEL,ACL_CATEGORY_CONNECTION,CLIENT_LIST_Keyspecs,0,NULL,2),.args=CLIENT_LIST_Args},
|
||||
{MAKE_CMD("no-evict","Sets the client eviction mode of the connection.","O(1)","7.0.0",CMD_DOC_NONE,NULL,NULL,"connection",COMMAND_GROUP_CONNECTION,CLIENT_NO_EVICT_History,0,CLIENT_NO_EVICT_Tips,0,clientCommand,3,CMD_ADMIN|CMD_NOSCRIPT|CMD_LOADING|CMD_STALE|CMD_SENTINEL,ACL_CATEGORY_CONNECTION,CLIENT_NO_EVICT_Keyspecs,0,NULL,1),.args=CLIENT_NO_EVICT_Args},
|
||||
{MAKE_CMD("no-touch","Controls whether commands sent by the client affect the LRU/LFU of accessed keys.","O(1)","7.2.0",CMD_DOC_NONE,NULL,NULL,"connection",COMMAND_GROUP_CONNECTION,CLIENT_NO_TOUCH_History,0,CLIENT_NO_TOUCH_Tips,0,clientCommand,3,CMD_NOSCRIPT|CMD_LOADING|CMD_STALE,ACL_CATEGORY_CONNECTION,CLIENT_NO_TOUCH_Keyspecs,0,NULL,1),.args=CLIENT_NO_TOUCH_Args},
|
||||
{MAKE_CMD("pause","Suspends commands processing.","O(1)","3.0.0",CMD_DOC_NONE,NULL,NULL,"connection",COMMAND_GROUP_CONNECTION,CLIENT_PAUSE_History,1,CLIENT_PAUSE_Tips,0,clientCommand,-3,CMD_ADMIN|CMD_NOSCRIPT|CMD_LOADING|CMD_STALE|CMD_SENTINEL,ACL_CATEGORY_CONNECTION,CLIENT_PAUSE_Keyspecs,0,NULL,2),.args=CLIENT_PAUSE_Args},
|
||||
|
|
|
|||
|
|
@ -31,6 +31,18 @@
|
|||
[
|
||||
"7.0.3",
|
||||
"Added `ssub` field."
|
||||
],
|
||||
[
|
||||
"7.2.0",
|
||||
"Added `lib-name` and `lib-ver` fields."
|
||||
],
|
||||
[
|
||||
"7.4.0",
|
||||
"Added `watch` field."
|
||||
],
|
||||
[
|
||||
"8.0.0",
|
||||
"Added `io-thread` field."
|
||||
]
|
||||
],
|
||||
"command_flags": [
|
||||
|
|
|
|||
12
src/config.c
12
src/config.c
|
|
@ -430,6 +430,7 @@ void loadServerConfigFromString(char *config) {
|
|||
{"list-max-ziplist-entries", 2, 2},
|
||||
{"list-max-ziplist-value", 2, 2},
|
||||
{"lua-replicate-commands", 2, 2},
|
||||
{"io-threads-do-reads", 2, 2},
|
||||
{NULL, 0},
|
||||
};
|
||||
char buf[1024];
|
||||
|
|
@ -2550,11 +2551,10 @@ static int updateMaxclients(const char **err) {
|
|||
*err = msg;
|
||||
return 0;
|
||||
}
|
||||
if ((unsigned int) aeGetSetSize(server.el) <
|
||||
server.maxclients + CONFIG_FDSET_INCR)
|
||||
{
|
||||
if (aeResizeSetSize(server.el,
|
||||
server.maxclients + CONFIG_FDSET_INCR) == AE_ERR)
|
||||
size_t newsize = server.maxclients + CONFIG_FDSET_INCR;
|
||||
if ((unsigned int) aeGetSetSize(server.el) < newsize) {
|
||||
if (aeResizeSetSize(server.el, newsize) == AE_ERR ||
|
||||
resizeAllIOThreadsEventLoops(newsize) == AE_ERR)
|
||||
{
|
||||
*err = "The event loop API used by Redis is not able to handle the specified number of clients";
|
||||
return 0;
|
||||
|
|
@ -3035,6 +3035,7 @@ static int applyClientMaxMemoryUsage(const char **err) {
|
|||
if (server.maxmemory_clients != 0)
|
||||
initServerClientMemUsageBuckets();
|
||||
|
||||
pauseAllIOThreads();
|
||||
/* When client eviction is enabled update memory buckets for all clients.
|
||||
* When disabled, clear that data structure. */
|
||||
listRewind(server.clients, &li);
|
||||
|
|
@ -3048,6 +3049,7 @@ static int applyClientMaxMemoryUsage(const char **err) {
|
|||
updateClientMemUsageAndBucket(c);
|
||||
}
|
||||
}
|
||||
resumeAllIOThreads();
|
||||
|
||||
if (server.maxmemory_clients == 0)
|
||||
freeServerClientMemUsageBuckets();
|
||||
|
|
|
|||
|
|
@ -47,6 +47,7 @@
|
|||
#define HAVE_PROC_SMAPS 1
|
||||
#define HAVE_PROC_SOMAXCONN 1
|
||||
#define HAVE_PROC_OOM_SCORE_ADJ 1
|
||||
#define HAVE_EVENT_FD 1
|
||||
#endif
|
||||
|
||||
/* Test for task_info() */
|
||||
|
|
|
|||
|
|
@ -156,14 +156,14 @@ void connTypeCleanupAll(void) {
|
|||
}
|
||||
|
||||
/* walk all the connection types until has pending data */
|
||||
int connTypeHasPendingData(void) {
|
||||
int connTypeHasPendingData(struct aeEventLoop *el) {
|
||||
ConnectionType *ct;
|
||||
int type;
|
||||
int ret = 0;
|
||||
|
||||
for (type = 0; type < CONN_TYPE_MAX; type++) {
|
||||
ct = connTypes[type];
|
||||
if (ct && ct->has_pending_data && (ret = ct->has_pending_data())) {
|
||||
if (ct && ct->has_pending_data && (ret = ct->has_pending_data(el))) {
|
||||
return ret;
|
||||
}
|
||||
}
|
||||
|
|
@ -172,7 +172,7 @@ int connTypeHasPendingData(void) {
|
|||
}
|
||||
|
||||
/* walk all the connection types and process pending data for each connection type */
|
||||
int connTypeProcessPendingData(void) {
|
||||
int connTypeProcessPendingData(struct aeEventLoop *el) {
|
||||
ConnectionType *ct;
|
||||
int type;
|
||||
int ret = 0;
|
||||
|
|
@ -180,7 +180,7 @@ int connTypeProcessPendingData(void) {
|
|||
for (type = 0; type < CONN_TYPE_MAX; type++) {
|
||||
ct = connTypes[type];
|
||||
if (ct && ct->process_pending_data) {
|
||||
ret += ct->process_pending_data();
|
||||
ret += ct->process_pending_data(el);
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -60,8 +60,8 @@ typedef struct ConnectionType {
|
|||
int (*listen)(connListener *listener);
|
||||
|
||||
/* create/shutdown/close connection */
|
||||
connection* (*conn_create)(void);
|
||||
connection* (*conn_create_accepted)(int fd, void *priv);
|
||||
connection* (*conn_create)(struct aeEventLoop *el);
|
||||
connection* (*conn_create_accepted)(struct aeEventLoop *el, int fd, void *priv);
|
||||
void (*shutdown)(struct connection *conn);
|
||||
void (*close)(struct connection *conn);
|
||||
|
||||
|
|
@ -81,9 +81,13 @@ typedef struct ConnectionType {
|
|||
ssize_t (*sync_read)(struct connection *conn, char *ptr, ssize_t size, long long timeout);
|
||||
ssize_t (*sync_readline)(struct connection *conn, char *ptr, ssize_t size, long long timeout);
|
||||
|
||||
/* event loop */
|
||||
void (*unbind_event_loop)(struct connection *conn);
|
||||
int (*rebind_event_loop)(struct connection *conn, aeEventLoop *el);
|
||||
|
||||
/* pending data */
|
||||
int (*has_pending_data)(void);
|
||||
int (*process_pending_data)(void);
|
||||
int (*has_pending_data)(struct aeEventLoop *el);
|
||||
int (*process_pending_data)(struct aeEventLoop *el);
|
||||
|
||||
/* TLS specified methods */
|
||||
sds (*get_peer_cert)(struct connection *conn);
|
||||
|
|
@ -98,6 +102,7 @@ struct connection {
|
|||
short int refs;
|
||||
unsigned short int iovcnt;
|
||||
void *private_data;
|
||||
struct aeEventLoop *el;
|
||||
ConnectionCallbackFunc conn_handler;
|
||||
ConnectionCallbackFunc write_handler;
|
||||
ConnectionCallbackFunc read_handler;
|
||||
|
|
@ -319,6 +324,28 @@ static inline int connHasReadHandler(connection *conn) {
|
|||
return conn->read_handler != NULL;
|
||||
}
|
||||
|
||||
/* Returns true if the connection is bound to an event loop */
|
||||
static inline int connHasEventLoop(connection *conn) {
|
||||
return conn->el != NULL;
|
||||
}
|
||||
|
||||
/* Unbind the current event loop from the connection, so that it can be
|
||||
* rebind to a different event loop in the future. */
|
||||
static inline void connUnbindEventLoop(connection *conn) {
|
||||
if (conn->el == NULL) return;
|
||||
connSetReadHandler(conn, NULL);
|
||||
connSetWriteHandler(conn, NULL);
|
||||
if (conn->type->unbind_event_loop)
|
||||
conn->type->unbind_event_loop(conn);
|
||||
conn->el = NULL;
|
||||
}
|
||||
|
||||
/* Rebind the connection to another event loop, read/write handlers must not
|
||||
* be installed in the current event loop */
|
||||
static inline int connRebindEventLoop(connection *conn, aeEventLoop *el) {
|
||||
return conn->type->rebind_event_loop(conn, el);
|
||||
}
|
||||
|
||||
/* Associate a private data pointer with the connection */
|
||||
static inline void connSetPrivateData(connection *conn, void *data) {
|
||||
conn->private_data = data;
|
||||
|
|
@ -379,14 +406,14 @@ ConnectionType *connectionTypeUnix(void);
|
|||
int connectionIndexByType(const char *typename);
|
||||
|
||||
/* Create a connection of specified type */
|
||||
static inline connection *connCreate(ConnectionType *ct) {
|
||||
return ct->conn_create();
|
||||
static inline connection *connCreate(struct aeEventLoop *el, ConnectionType *ct) {
|
||||
return ct->conn_create(el);
|
||||
}
|
||||
|
||||
/* Create an accepted connection of specified type.
|
||||
* priv is connection type specified argument */
|
||||
static inline connection *connCreateAccepted(ConnectionType *ct, int fd, void *priv) {
|
||||
return ct->conn_create_accepted(fd, priv);
|
||||
static inline connection *connCreateAccepted(struct aeEventLoop *el, ConnectionType *ct, int fd, void *priv) {
|
||||
return ct->conn_create_accepted(el, fd, priv);
|
||||
}
|
||||
|
||||
/* Configure a connection type. A typical case is to configure TLS.
|
||||
|
|
@ -400,10 +427,10 @@ static inline int connTypeConfigure(ConnectionType *ct, void *priv, int reconfig
|
|||
void connTypeCleanupAll(void);
|
||||
|
||||
/* Test all the connection type has pending data or not. */
|
||||
int connTypeHasPendingData(void);
|
||||
int connTypeHasPendingData(struct aeEventLoop *el);
|
||||
|
||||
/* walk all the connection types and process pending data for each connection type */
|
||||
int connTypeProcessPendingData(void);
|
||||
int connTypeProcessPendingData(struct aeEventLoop *el);
|
||||
|
||||
/* Listen on an initialized listener */
|
||||
static inline int connListen(connListener *listener) {
|
||||
|
|
|
|||
|
|
@ -2451,6 +2451,8 @@ void removeSigSegvHandlers(void) {
|
|||
}
|
||||
|
||||
void printCrashReport(void) {
|
||||
server.crashing = 1;
|
||||
|
||||
/* Log INFO and CLIENT LIST */
|
||||
logServerInfo();
|
||||
|
||||
|
|
|
|||
|
|
@ -0,0 +1,97 @@
|
|||
/* eventnotifier.c -- An event notifier based on eventfd or pipe.
|
||||
*
|
||||
* Copyright (c) 2024-Present, Redis Ltd.
|
||||
* All rights reserved.
|
||||
*
|
||||
* Licensed under your choice of the Redis Source Available License 2.0
|
||||
* (RSALv2) or the Server Side Public License v1 (SSPLv1).
|
||||
*/
|
||||
|
||||
#include "eventnotifier.h"
|
||||
|
||||
#include <stdint.h>
|
||||
#include <unistd.h>
|
||||
#include <fcntl.h>
|
||||
#ifdef HAVE_EVENT_FD
|
||||
#include <sys/eventfd.h>
|
||||
#endif
|
||||
|
||||
#include "anet.h"
|
||||
#include "zmalloc.h"
|
||||
|
||||
eventNotifier* createEventNotifier(void) {
|
||||
eventNotifier *en = zmalloc(sizeof(eventNotifier));
|
||||
if (!en) return NULL;
|
||||
|
||||
#ifdef HAVE_EVENT_FD
|
||||
if ((en->efd = eventfd(0, EFD_NONBLOCK| EFD_CLOEXEC)) != -1) {
|
||||
return en;
|
||||
}
|
||||
#else
|
||||
if (anetPipe(en->pipefd, O_CLOEXEC|O_NONBLOCK, O_CLOEXEC|O_NONBLOCK) != -1) {
|
||||
return en;
|
||||
}
|
||||
#endif
|
||||
|
||||
/* Clean up if error. */
|
||||
zfree(en);
|
||||
return NULL;
|
||||
}
|
||||
|
||||
int getReadEventFd(struct eventNotifier *en) {
|
||||
#ifdef HAVE_EVENT_FD
|
||||
return en->efd;
|
||||
#else
|
||||
return en->pipefd[0];
|
||||
#endif
|
||||
}
|
||||
|
||||
int getWriteEventFd(struct eventNotifier *en) {
|
||||
#ifdef HAVE_EVENT_FD
|
||||
return en->efd;
|
||||
#else
|
||||
return en->pipefd[1];
|
||||
#endif
|
||||
}
|
||||
|
||||
int triggerEventNotifier(struct eventNotifier *en) {
|
||||
#ifdef HAVE_EVENT_FD
|
||||
uint64_t u = 1;
|
||||
if (write(en->efd, &u, sizeof(uint64_t)) == -1) {
|
||||
return EN_ERR;
|
||||
}
|
||||
#else
|
||||
char buf[1] = {'R'};
|
||||
if (write(en->pipefd[1], buf, 1) == -1) {
|
||||
return EN_ERR;
|
||||
}
|
||||
#endif
|
||||
return EN_OK;
|
||||
}
|
||||
|
||||
int handleEventNotifier(struct eventNotifier *en) {
|
||||
#ifdef HAVE_EVENT_FD
|
||||
uint64_t u;
|
||||
if (read(en->efd, &u, sizeof(uint64_t)) == -1) {
|
||||
return EN_ERR;
|
||||
}
|
||||
#else
|
||||
char buf[1];
|
||||
if (read(en->pipefd[0], buf, 1) == -1) {
|
||||
return EN_ERR;
|
||||
}
|
||||
#endif
|
||||
return EN_OK;
|
||||
}
|
||||
|
||||
void freeEventNotifier(struct eventNotifier *en) {
|
||||
#ifdef HAVE_EVENT_FD
|
||||
close(en->efd);
|
||||
#else
|
||||
close(en->pipefd[0]);
|
||||
close(en->pipefd[1]);
|
||||
#endif
|
||||
|
||||
/* Free memory */
|
||||
zfree(en);
|
||||
}
|
||||
|
|
@ -0,0 +1,33 @@
|
|||
/* eventnotifier.h -- An event notifier based on eventfd or pipe.
|
||||
*
|
||||
* Copyright (c) 2024-Present, Redis Ltd.
|
||||
* All rights reserved.
|
||||
*
|
||||
* Licensed under your choice of the Redis Source Available License 2.0
|
||||
* (RSALv2) or the Server Side Public License v1 (SSPLv1).
|
||||
*/
|
||||
|
||||
#ifndef EVENTNOTIFIER_H
|
||||
#define EVENTNOTIFIER_H
|
||||
|
||||
#include "config.h"
|
||||
|
||||
#define EN_OK 0
|
||||
#define EN_ERR -1
|
||||
|
||||
typedef struct eventNotifier {
|
||||
#ifdef HAVE_EVENT_FD
|
||||
int efd;
|
||||
#else
|
||||
int pipefd[2];
|
||||
#endif
|
||||
} eventNotifier;
|
||||
|
||||
eventNotifier* createEventNotifier(void);
|
||||
int getReadEventFd(struct eventNotifier *en);
|
||||
int getWriteEventFd(struct eventNotifier *en);
|
||||
int triggerEventNotifier(struct eventNotifier *en);
|
||||
int handleEventNotifier(struct eventNotifier *en);
|
||||
void freeEventNotifier(struct eventNotifier *en);
|
||||
|
||||
#endif
|
||||
|
|
@ -0,0 +1,631 @@
|
|||
/* iothread.c -- The threaded io implementation.
|
||||
*
|
||||
* Copyright (c) 2024-Present, Redis Ltd.
|
||||
* All rights reserved.
|
||||
*
|
||||
* Licensed under your choice of the Redis Source Available License 2.0
|
||||
* (RSALv2) or the Server Side Public License v1 (SSPLv1).
|
||||
*/
|
||||
|
||||
#include "server.h"
|
||||
|
||||
/* IO threads. */
|
||||
static IOThread IOThreads[IO_THREADS_MAX_NUM];
|
||||
|
||||
/* For main thread */
|
||||
static list *mainThreadPendingClientsToIOThreads[IO_THREADS_MAX_NUM]; /* Clients to IO threads */
|
||||
static list *mainThreadProcessingClients[IO_THREADS_MAX_NUM]; /* Clients in processing */
|
||||
static list *mainThreadPendingClients[IO_THREADS_MAX_NUM]; /* Pending clients from IO threads */
|
||||
static pthread_mutex_t mainThreadPendingClientsMutexes[IO_THREADS_MAX_NUM]; /* Mutex for pending clients */
|
||||
static eventNotifier* mainThreadPendingClientsNotifiers[IO_THREADS_MAX_NUM]; /* Notifier for pending clients */
|
||||
|
||||
/* When IO threads read a complete query of clients or want to free clients, it
|
||||
* should remove it from its clients list and put the client in the list to main
|
||||
* thread, we will send these clients to main thread in IOThreadBeforeSleep. */
|
||||
void enqueuePendingClientsToMainThread(client *c, int unbind) {
|
||||
/* If the IO thread may no longer manage it, such as closing client, we should
|
||||
* unbind client from event loop, so main thread doesn't need to do it costly. */
|
||||
if (unbind) connUnbindEventLoop(c->conn);
|
||||
/* Just skip if it already is transferred. */
|
||||
if (c->io_thread_client_list_node) {
|
||||
listDelNode(IOThreads[c->tid].clients, c->io_thread_client_list_node);
|
||||
c->io_thread_client_list_node = NULL;
|
||||
/* Disable read and write to avoid race when main thread processes. */
|
||||
c->io_flags &= ~(CLIENT_IO_READ_ENABLED | CLIENT_IO_WRITE_ENABLED);
|
||||
listAddNodeTail(IOThreads[c->tid].pending_clients_to_main_thread, c);
|
||||
}
|
||||
}
|
||||
|
||||
/* Unbind connection of client from io thread event loop, write and read handlers
|
||||
* also be removed, ensures that we can operate the client safely. */
|
||||
void unbindClientFromIOThreadEventLoop(client *c) {
|
||||
serverAssert(c->tid != IOTHREAD_MAIN_THREAD_ID &&
|
||||
c->running_tid == IOTHREAD_MAIN_THREAD_ID);
|
||||
if (!connHasEventLoop(c->conn)) return;
|
||||
/* As calling in main thread, we should pause the io thread to make it safe. */
|
||||
pauseIOThread(c->tid);
|
||||
connUnbindEventLoop(c->conn);
|
||||
resumeIOThread(c->tid);
|
||||
}
|
||||
|
||||
/* When main thread is processing a client from IO thread, and wants to keep it,
|
||||
* we should unbind connection of client from io thread event loop first,
|
||||
* and then bind the client connection into server's event loop. */
|
||||
void keepClientInMainThread(client *c) {
|
||||
serverAssert(c->tid != IOTHREAD_MAIN_THREAD_ID &&
|
||||
c->running_tid == IOTHREAD_MAIN_THREAD_ID);
|
||||
/* IO thread no longer manage it. */
|
||||
server.io_threads_clients_num[c->tid]--;
|
||||
/* Unbind connection of client from io thread event loop. */
|
||||
unbindClientFromIOThreadEventLoop(c);
|
||||
/* Let main thread to run it, rebind event loop and read handler */
|
||||
connRebindEventLoop(c->conn, server.el);
|
||||
connSetReadHandler(c->conn, readQueryFromClient);
|
||||
c->io_flags |= CLIENT_IO_READ_ENABLED | CLIENT_IO_WRITE_ENABLED;
|
||||
c->running_tid = IOTHREAD_MAIN_THREAD_ID;
|
||||
c->tid = IOTHREAD_MAIN_THREAD_ID;
|
||||
/* Main thread starts to manage it. */
|
||||
server.io_threads_clients_num[c->tid]++;
|
||||
}
|
||||
|
||||
/* If the client is managed by IO thread, we should fetch it from IO thread
|
||||
* and then main thread will can process it. Just like IO Thread transfers
|
||||
* the client to the main thread for processing. */
|
||||
void fetchClientFromIOThread(client *c) {
|
||||
serverAssert(c->tid != IOTHREAD_MAIN_THREAD_ID &&
|
||||
c->running_tid != IOTHREAD_MAIN_THREAD_ID);
|
||||
pauseIOThread(c->tid);
|
||||
/* Remove the client from clients list of IO thread or main thread. */
|
||||
if (c->io_thread_client_list_node) {
|
||||
listDelNode(IOThreads[c->tid].clients, c->io_thread_client_list_node);
|
||||
c->io_thread_client_list_node = NULL;
|
||||
} else {
|
||||
list *clients[5] = {
|
||||
IOThreads[c->tid].pending_clients,
|
||||
IOThreads[c->tid].pending_clients_to_main_thread,
|
||||
mainThreadPendingClients[c->tid],
|
||||
mainThreadProcessingClients[c->tid],
|
||||
mainThreadPendingClientsToIOThreads[c->tid]
|
||||
};
|
||||
for (int i = 0; i < 5; i++) {
|
||||
listNode *ln = listSearchKey(clients[i], c);
|
||||
if (ln) {
|
||||
listDelNode(clients[i], ln);
|
||||
/* Client only can be in one client list. */
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
/* Unbind connection of client from io thread event loop. */
|
||||
connUnbindEventLoop(c->conn);
|
||||
/* Now main thread can process it. */
|
||||
c->running_tid = IOTHREAD_MAIN_THREAD_ID;
|
||||
resumeIOThread(c->tid);
|
||||
}
|
||||
|
||||
/* For some clients, we must handle them in the main thread, since there is
|
||||
* data race to be processed in IO threads.
|
||||
*
|
||||
* - Close ASAP, we must free the client in main thread.
|
||||
* - Replica, pubsub, monitor, blocked, tracking clients, main thread may
|
||||
* directly write them a reply when conditions are met.
|
||||
* - Script command with debug may operate connection directly. */
|
||||
int isClientMustHandledByMainThread(client *c) {
|
||||
if (c->flags & (CLIENT_CLOSE_ASAP | CLIENT_MASTER | CLIENT_SLAVE |
|
||||
CLIENT_PUBSUB | CLIENT_MONITOR | CLIENT_BLOCKED |
|
||||
CLIENT_UNBLOCKED | CLIENT_TRACKING | CLIENT_LUA_DEBUG |
|
||||
CLIENT_LUA_DEBUG_SYNC))
|
||||
{
|
||||
return 1;
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
/* When the main thread accepts a new client or transfers clients to IO threads,
|
||||
* it assigns the client to the IO thread with the fewest clients. */
|
||||
void assignClientToIOThread(client *c) {
|
||||
serverAssert(c->tid == IOTHREAD_MAIN_THREAD_ID);
|
||||
/* Find the IO thread with the fewest clients. */
|
||||
int min_id = 0;
|
||||
int min = INT_MAX;
|
||||
for (int i = 1; i < server.io_threads_num; i++) {
|
||||
if (server.io_threads_clients_num[i] < min) {
|
||||
min = server.io_threads_clients_num[i];
|
||||
min_id = i;
|
||||
}
|
||||
}
|
||||
|
||||
/* Assign the client to the IO thread. */
|
||||
server.io_threads_clients_num[c->tid]--;
|
||||
c->tid = min_id;
|
||||
c->running_tid = min_id;
|
||||
server.io_threads_clients_num[min_id]++;
|
||||
|
||||
/* Unbind connection of client from main thread event loop, disable read and
|
||||
* write, and then put it in the list, main thread will send these clients
|
||||
* to IO thread in beforeSleep. */
|
||||
connUnbindEventLoop(c->conn);
|
||||
c->io_flags &= ~(CLIENT_IO_READ_ENABLED | CLIENT_IO_WRITE_ENABLED);
|
||||
listAddNodeTail(mainThreadPendingClientsToIOThreads[c->tid], c);
|
||||
}
|
||||
|
||||
/* If updating maxclients config, we not only resize the event loop of main thread
|
||||
* but also resize the event loop of all io threads, and if one thread is failed,
|
||||
* it is failed totally, since a fd can be distributed into any IO thread. */
|
||||
int resizeAllIOThreadsEventLoops(size_t newsize) {
|
||||
int result = AE_OK;
|
||||
if (server.io_threads_num <= 1) return result;
|
||||
|
||||
/* To make context safe. */
|
||||
pauseAllIOThreads();
|
||||
for (int i = 1; i < server.io_threads_num; i++) {
|
||||
IOThread *t = &IOThreads[i];
|
||||
if (aeResizeSetSize(t->el, newsize) == AE_ERR)
|
||||
result = AE_ERR;
|
||||
}
|
||||
resumeAllIOThreads();
|
||||
return result;
|
||||
}
|
||||
|
||||
/* In the main thread, we may want to operate data of io threads, maybe uninstall
|
||||
* event handler, access query/output buffer or resize event loop, we need a clean
|
||||
* and safe context to do that. We pause io thread in IOThreadBeforeSleep, do some
|
||||
* jobs and then resume it. To avoid thread suspended, we use busy waiting to confirm
|
||||
* the target status. Besides we use atomic variable to make sure memory visibility
|
||||
* and ordering.
|
||||
*
|
||||
* Make sure that only the main thread can call these function,
|
||||
* - pauseIOThread, resumeIOThread
|
||||
* - pauseAllIOThreads, resumeAllIOThreads
|
||||
* - pauseIOThreadsRange, resumeIOThreadsRange
|
||||
*
|
||||
* The main thread will pause the io thread, and then wait for the io thread to
|
||||
* be paused. The io thread will check the paused status in IOThreadBeforeSleep,
|
||||
* and then pause itself.
|
||||
*
|
||||
* The main thread will resume the io thread, and then wait for the io thread to
|
||||
* be resumed. The io thread will check the paused status in IOThreadBeforeSleep,
|
||||
* and then resume itself.
|
||||
*/
|
||||
|
||||
/* We may pause the same io thread nestedly, so we need to record the times of
|
||||
* pausing, and only when the times of pausing is 0, we can pause the io thread,
|
||||
* and only when the times of pausing is 1, we can resume the io thread. */
|
||||
static int PausedIOThreads[IO_THREADS_MAX_NUM] = {0};
|
||||
|
||||
/* Pause the specific range of io threads, and wait for them to be paused. */
|
||||
void pauseIOThreadsRange(int start, int end) {
|
||||
if (server.io_threads_num <= 1) return;
|
||||
serverAssert(start >= 1 && end < server.io_threads_num && start <= end);
|
||||
serverAssert(pthread_equal(pthread_self(), server.main_thread_id));
|
||||
|
||||
/* Try to make all io threads paused in parallel */
|
||||
for (int i = start; i <= end; i++) {
|
||||
PausedIOThreads[i]++;
|
||||
/* Skip if already paused */
|
||||
if (PausedIOThreads[i] > 1) continue;
|
||||
|
||||
int paused;
|
||||
atomicGetWithSync(IOThreads[i].paused, paused);
|
||||
/* Don't support to call reentrant */
|
||||
serverAssert(paused == IO_THREAD_UNPAUSED);
|
||||
atomicSetWithSync(IOThreads[i].paused, IO_THREAD_PAUSING);
|
||||
/* Just notify io thread, no actual job, since io threads check paused
|
||||
* status in IOThreadBeforeSleep, so just wake it up if polling wait. */
|
||||
triggerEventNotifier(IOThreads[i].pending_clients_notifier);
|
||||
}
|
||||
|
||||
/* Wait for all io threads paused */
|
||||
for (int i = start; i <= end; i++) {
|
||||
if (PausedIOThreads[i] > 1) continue;
|
||||
int paused = IO_THREAD_PAUSING;
|
||||
while (paused != IO_THREAD_PAUSED) {
|
||||
atomicGetWithSync(IOThreads[i].paused, paused);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/* Resume the specific range of io threads, and wait for them to be resumed. */
|
||||
void resumeIOThreadsRange(int start, int end) {
|
||||
if (server.io_threads_num <= 1) return;
|
||||
serverAssert(start >= 1 && end < server.io_threads_num && start <= end);
|
||||
serverAssert(pthread_equal(pthread_self(), server.main_thread_id));
|
||||
|
||||
for (int i = start; i <= end; i++) {
|
||||
serverAssert(PausedIOThreads[i] > 0);
|
||||
PausedIOThreads[i]--;
|
||||
if (PausedIOThreads[i] > 0) continue;
|
||||
|
||||
int paused;
|
||||
/* Check if it is paused, since we must call 'pause' and
|
||||
* 'resume' in pairs */
|
||||
atomicGetWithSync(IOThreads[i].paused, paused);
|
||||
serverAssert(paused == IO_THREAD_PAUSED);
|
||||
/* Resume */
|
||||
atomicSetWithSync(IOThreads[i].paused, IO_THREAD_RESUMING);
|
||||
while (paused != IO_THREAD_UNPAUSED) {
|
||||
atomicGetWithSync(IOThreads[i].paused, paused);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/* The IO thread checks whether it is being paused, and if so, it pauses itself
|
||||
* and waits for resuming, corresponding to the pause/resumeIOThread* functions.
|
||||
* Currently, this is only called in IOThreadBeforeSleep, as there are no pending
|
||||
* I/O events at this point, with a clean context. */
|
||||
void handlePauseAndResume(IOThread *t) {
|
||||
int paused;
|
||||
/* Check if i am being paused. */
|
||||
atomicGetWithSync(t->paused, paused);
|
||||
if (paused == IO_THREAD_PAUSING) {
|
||||
atomicSetWithSync(t->paused, IO_THREAD_PAUSED);
|
||||
/* Wait for resuming */
|
||||
while (paused != IO_THREAD_RESUMING) {
|
||||
atomicGetWithSync(t->paused, paused);
|
||||
}
|
||||
atomicSetWithSync(t->paused, IO_THREAD_UNPAUSED);
|
||||
}
|
||||
}
|
||||
|
||||
/* Pause the specific io thread, and wait for it to be paused. */
|
||||
void pauseIOThread(int id) {
|
||||
pauseIOThreadsRange(id, id);
|
||||
}
|
||||
|
||||
/* Resume the specific io thread, and wait for it to be resumed. */
|
||||
void resumeIOThread(int id) {
|
||||
resumeIOThreadsRange(id, id);
|
||||
}
|
||||
|
||||
/* Pause all io threads, and wait for them to be paused. */
|
||||
void pauseAllIOThreads(void) {
|
||||
pauseIOThreadsRange(1, server.io_threads_num-1);
|
||||
}
|
||||
|
||||
/* Resume all io threads, and wait for them to be resumed. */
|
||||
void resumeAllIOThreads(void) {
|
||||
resumeIOThreadsRange(1, server.io_threads_num-1);
|
||||
}
|
||||
|
||||
/* Add the pending clients to the list of IO threads, and trigger an event to
|
||||
* notify io threads to handle. */
|
||||
int sendPendingClientsToIOThreads(void) {
|
||||
int processed = 0;
|
||||
for (int i = 1; i < server.io_threads_num; i++) {
|
||||
int len = listLength(mainThreadPendingClientsToIOThreads[i]);
|
||||
if (len > 0) {
|
||||
IOThread *t = &IOThreads[i];
|
||||
pthread_mutex_lock(&t->pending_clients_mutex);
|
||||
listJoin(t->pending_clients, mainThreadPendingClientsToIOThreads[i]);
|
||||
pthread_mutex_unlock(&t->pending_clients_mutex);
|
||||
/* Trigger an event, maybe an error is returned when buffer is full
|
||||
* if using pipe, but no worry, io thread will handle all clients
|
||||
* in list when receiving a notification. */
|
||||
triggerEventNotifier(t->pending_clients_notifier);
|
||||
}
|
||||
processed += len;
|
||||
}
|
||||
return processed;
|
||||
}
|
||||
|
||||
extern int ProcessingEventsWhileBlocked;
|
||||
|
||||
/* The main thread processes the clients from IO threads, these clients may have
|
||||
* a complete command to execute or need to be freed. Note that IO threads never
|
||||
* free client since this operation access much server data.
|
||||
*
|
||||
* Please notice that this function may be called reentrantly, i,e, the same goes
|
||||
* for handleClientsFromIOThread and processClientsOfAllIOThreads. For example,
|
||||
* when processing script command, it may call processEventsWhileBlocked to
|
||||
* process new events, if the clients with fired events from the same io thread,
|
||||
* it may call this function reentrantly. */
|
||||
void processClientsFromIOThread(IOThread *t) {
|
||||
listNode *node = NULL;
|
||||
|
||||
while (listLength(mainThreadProcessingClients[t->id])) {
|
||||
/* Each time we pop up only the first client to process to guarantee
|
||||
* reentrancy safety. */
|
||||
if (node) zfree(node);
|
||||
node = listFirst(mainThreadProcessingClients[t->id]);
|
||||
listUnlinkNode(mainThreadProcessingClients[t->id], node);
|
||||
client *c = listNodeValue(node);
|
||||
|
||||
/* Make sure the client is readable or writable in io thread to
|
||||
* avoid data race. */
|
||||
serverAssert(!(c->io_flags & (CLIENT_IO_READ_ENABLED | CLIENT_IO_WRITE_ENABLED)));
|
||||
serverAssert(!(c->flags & CLIENT_CLOSE_ASAP));
|
||||
|
||||
/* Let main thread to run it, set running thread id first. */
|
||||
c->running_tid = IOTHREAD_MAIN_THREAD_ID;
|
||||
|
||||
/* If a read error occurs, handle it in the main thread first, since we
|
||||
* want to print logs about client information before freeing. */
|
||||
if (c->read_error) handleClientReadError(c);
|
||||
|
||||
/* The client is asked to close in IO thread. */
|
||||
if (c->io_flags & CLIENT_IO_CLOSE_ASAP) {
|
||||
freeClient(c);
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Update the client in the mem usage */
|
||||
updateClientMemUsageAndBucket(c);
|
||||
|
||||
/* Process the pending command and input buffer. */
|
||||
if (!c->read_error && c->io_flags & CLIENT_IO_PENDING_COMMAND) {
|
||||
c->flags |= CLIENT_PENDING_COMMAND;
|
||||
if (processPendingCommandAndInputBuffer(c) == C_ERR) {
|
||||
/* If the client is no longer valid, it must be freed safely. */
|
||||
continue;
|
||||
}
|
||||
}
|
||||
|
||||
/* We may have pending replies if io thread may not finish writing
|
||||
* reply to client, so we did not put the client in pending write
|
||||
* queue. And we should do that first since we may keep the client
|
||||
* in main thread instead of returning to io threads. */
|
||||
if (!(c->flags & CLIENT_PENDING_WRITE) && clientHasPendingReplies(c))
|
||||
putClientInPendingWriteQueue(c);
|
||||
|
||||
/* The client only can be processed in the main thread, otherwise data
|
||||
* race will happen, since we may touch client's data in main thread. */
|
||||
if (isClientMustHandledByMainThread(c)) {
|
||||
keepClientInMainThread(c);
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Remove this client from pending write clients queue of main thread,
|
||||
* And some clients may do not have reply if CLIENT REPLY OFF/SKIP. */
|
||||
if (c->flags & CLIENT_PENDING_WRITE) {
|
||||
c->flags &= ~CLIENT_PENDING_WRITE;
|
||||
listUnlinkNode(server.clients_pending_write, &c->clients_pending_write_node);
|
||||
}
|
||||
c->running_tid = c->tid;
|
||||
listLinkNodeHead(mainThreadPendingClientsToIOThreads[c->tid], node);
|
||||
node = NULL;
|
||||
}
|
||||
if (node) zfree(node);
|
||||
|
||||
/* Trigger the io thread to handle these clients ASAP to make them processed
|
||||
* in parallel.
|
||||
*
|
||||
* If AOF fsync policy is always, we should not let io thread handle these
|
||||
* clients now since we don't flush AOF buffer to file and sync yet.
|
||||
* So these clients will be delayed to send io threads in beforeSleep after
|
||||
* flushAppendOnlyFile.
|
||||
*
|
||||
* If we are in processEventsWhileBlocked, we don't send clients to io threads
|
||||
* now, we want to update server.events_processed_while_blocked accurately. */
|
||||
if (listLength(mainThreadPendingClientsToIOThreads[t->id]) &&
|
||||
server.aof_fsync != AOF_FSYNC_ALWAYS &&
|
||||
!ProcessingEventsWhileBlocked)
|
||||
{
|
||||
pthread_mutex_lock(&(t->pending_clients_mutex));
|
||||
listJoin(t->pending_clients, mainThreadPendingClientsToIOThreads[t->id]);
|
||||
pthread_mutex_unlock(&(t->pending_clients_mutex));
|
||||
triggerEventNotifier(t->pending_clients_notifier);
|
||||
}
|
||||
}
|
||||
|
||||
/* When the io thread finishes processing the client with the read event, it will
|
||||
* notify the main thread through event triggering in IOThreadBeforeSleep. The main
|
||||
* thread handles the event through this function. */
|
||||
void handleClientsFromIOThread(struct aeEventLoop *el, int fd, void *ptr, int mask) {
|
||||
UNUSED(el);
|
||||
UNUSED(mask);
|
||||
|
||||
IOThread *t = ptr;
|
||||
|
||||
/* Handle fd event first. */
|
||||
serverAssert(fd == getReadEventFd(mainThreadPendingClientsNotifiers[t->id]));
|
||||
handleEventNotifier(mainThreadPendingClientsNotifiers[t->id]);
|
||||
|
||||
/* Get the list of clients to process. */
|
||||
pthread_mutex_lock(&mainThreadPendingClientsMutexes[t->id]);
|
||||
listJoin(mainThreadProcessingClients[t->id], mainThreadPendingClients[t->id]);
|
||||
pthread_mutex_unlock(&mainThreadPendingClientsMutexes[t->id]);
|
||||
if (listLength(mainThreadProcessingClients[t->id]) == 0) return;
|
||||
|
||||
/* Process the clients from IO threads. */
|
||||
processClientsFromIOThread(t);
|
||||
}
|
||||
|
||||
/* In the new threaded io design, one thread may process multiple clients, so when
|
||||
* an io thread notifies the main thread of an event, there may be multiple clients
|
||||
* with commands that need to be processed. But in the event handler function
|
||||
* handleClientsFromIOThread may be blocked when processing the specific command,
|
||||
* the previous clients can not get a reply, and the subsequent clients can not be
|
||||
* processed, so we need to handle this scenario in beforeSleep. The function is to
|
||||
* process the commands of subsequent clients from io threads. And another function
|
||||
* sendPendingClientsToIOThreads make sure clients from io thread can get replies.
|
||||
* See also beforeSleep. */
|
||||
void processClientsOfAllIOThreads(void) {
|
||||
for (int i = 1; i < server.io_threads_num; i++) {
|
||||
processClientsFromIOThread(&IOThreads[i]);
|
||||
}
|
||||
}
|
||||
|
||||
/* After the main thread processes the clients, it will send the clients back to
|
||||
* io threads to handle, and fire an event, the io thread handles the event by
|
||||
* this function. If the client is not binded to the event loop, we should bind
|
||||
* it first and install read handler, and we don't uninstall client read handler
|
||||
* unless freeing client. If the client has pending reply, we just reply to client
|
||||
* first, and then install write handler if needed. */
|
||||
void handleClientsFromMainThread(struct aeEventLoop *ae, int fd, void *ptr, int mask) {
|
||||
UNUSED(ae);
|
||||
UNUSED(mask);
|
||||
|
||||
IOThread *t = ptr;
|
||||
|
||||
/* Handle fd event first. */
|
||||
serverAssert(fd == getReadEventFd(t->pending_clients_notifier));
|
||||
handleEventNotifier(t->pending_clients_notifier);
|
||||
|
||||
pthread_mutex_lock(&t->pending_clients_mutex);
|
||||
listJoin(t->processing_clients, t->pending_clients);
|
||||
pthread_mutex_unlock(&t->pending_clients_mutex);
|
||||
if (listLength(t->processing_clients) == 0) return;
|
||||
|
||||
listIter li;
|
||||
listNode *ln;
|
||||
listRewind(t->processing_clients, &li);
|
||||
while((ln = listNext(&li))) {
|
||||
client *c = listNodeValue(ln);
|
||||
serverAssert(!(c->io_flags & (CLIENT_IO_READ_ENABLED | CLIENT_IO_WRITE_ENABLED)));
|
||||
/* Main thread must handle clients with CLIENT_CLOSE_ASAP flag, since
|
||||
* we only set io_flags when clients in io thread are freed ASAP. */
|
||||
serverAssert(!(c->flags & CLIENT_CLOSE_ASAP));
|
||||
|
||||
/* Link client in IO thread clients list first. */
|
||||
serverAssert(c->io_thread_client_list_node == NULL);
|
||||
listAddNodeTail(t->clients, c);
|
||||
c->io_thread_client_list_node = listLast(t->clients);
|
||||
|
||||
/* The client is asked to close, we just let main thread free it. */
|
||||
if (c->io_flags & CLIENT_IO_CLOSE_ASAP) {
|
||||
enqueuePendingClientsToMainThread(c, 1);
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Enable read and write and reset some flags. */
|
||||
c->io_flags |= CLIENT_IO_READ_ENABLED | CLIENT_IO_WRITE_ENABLED;
|
||||
c->io_flags &= ~CLIENT_IO_PENDING_COMMAND;
|
||||
|
||||
/* Only bind once, we never remove read handler unless freeing client. */
|
||||
if (!connHasEventLoop(c->conn)) {
|
||||
connRebindEventLoop(c->conn, t->el);
|
||||
serverAssert(!connHasReadHandler(c->conn));
|
||||
connSetReadHandler(c->conn, readQueryFromClient);
|
||||
}
|
||||
|
||||
/* If the client has pending replies, write replies to client. */
|
||||
if (clientHasPendingReplies(c)) {
|
||||
writeToClient(c, 0);
|
||||
if (!(c->io_flags & CLIENT_IO_CLOSE_ASAP) && clientHasPendingReplies(c)) {
|
||||
connSetWriteHandler(c->conn, sendReplyToClient);
|
||||
}
|
||||
}
|
||||
}
|
||||
listEmpty(t->processing_clients);
|
||||
}
|
||||
|
||||
void IOThreadBeforeSleep(struct aeEventLoop *el) {
|
||||
IOThread *t = el->privdata[0];
|
||||
|
||||
/* Handle pending data(typical TLS). */
|
||||
connTypeProcessPendingData(el);
|
||||
|
||||
/* If any connection type(typical TLS) still has pending unread data don't sleep at all. */
|
||||
aeSetDontWait(el, connTypeHasPendingData(el));
|
||||
|
||||
/* Check if i am being paused, pause myself and resume. */
|
||||
handlePauseAndResume(t);
|
||||
|
||||
/* Check if there are clients to be processed in main thread, and then join
|
||||
* them to the list of main thread. */
|
||||
if (listLength(t->pending_clients_to_main_thread) > 0) {
|
||||
pthread_mutex_lock(&mainThreadPendingClientsMutexes[t->id]);
|
||||
listJoin(mainThreadPendingClients[t->id], t->pending_clients_to_main_thread);
|
||||
pthread_mutex_unlock(&mainThreadPendingClientsMutexes[t->id]);
|
||||
/* Trigger an event, maybe an error is returned when buffer is full
|
||||
* if using pipe, but no worry, main thread will handle all clients
|
||||
* in list when receiving a notification. */
|
||||
triggerEventNotifier(mainThreadPendingClientsNotifiers[t->id]);
|
||||
}
|
||||
}
|
||||
|
||||
/* The main function of IO thread, it will run an event loop. The mian thread
|
||||
* and IO thread will communicate through event notifier. */
|
||||
void *IOThreadMain(void *ptr) {
|
||||
IOThread *t = ptr;
|
||||
char thdname[16];
|
||||
snprintf(thdname, sizeof(thdname), "io_thd_%d", t->id);
|
||||
redis_set_thread_title(thdname);
|
||||
redisSetCpuAffinity(server.server_cpulist);
|
||||
makeThreadKillable();
|
||||
aeSetBeforeSleepProc(t->el, IOThreadBeforeSleep);
|
||||
aeMain(t->el);
|
||||
return NULL;
|
||||
}
|
||||
|
||||
/* Initialize the data structures needed for threaded I/O. */
|
||||
void initThreadedIO(void) {
|
||||
if (server.io_threads_num <= 1) return;
|
||||
|
||||
server.io_threads_active = 1;
|
||||
|
||||
if (server.io_threads_num > IO_THREADS_MAX_NUM) {
|
||||
serverLog(LL_WARNING,"Fatal: too many I/O threads configured. "
|
||||
"The maximum number is %d.", IO_THREADS_MAX_NUM);
|
||||
exit(1);
|
||||
}
|
||||
|
||||
/* Spawn and initialize the I/O threads. */
|
||||
for (int i = 1; i < server.io_threads_num; i++) {
|
||||
IOThread *t = &IOThreads[i];
|
||||
t->id = i;
|
||||
t->el = aeCreateEventLoop(server.maxclients+CONFIG_FDSET_INCR);
|
||||
t->el->privdata[0] = t;
|
||||
t->pending_clients = listCreate();
|
||||
t->processing_clients = listCreate();
|
||||
t->pending_clients_to_main_thread = listCreate();
|
||||
t->clients = listCreate();
|
||||
atomicSetWithSync(t->paused, IO_THREAD_UNPAUSED);
|
||||
|
||||
pthread_mutexattr_t *attr = NULL;
|
||||
#if defined(__linux__) && defined(__GLIBC__)
|
||||
attr = zmalloc(sizeof(pthread_mutexattr_t));
|
||||
pthread_mutexattr_init(attr);
|
||||
pthread_mutexattr_settype(attr, PTHREAD_MUTEX_ADAPTIVE_NP);
|
||||
#endif
|
||||
pthread_mutex_init(&t->pending_clients_mutex, attr);
|
||||
|
||||
t->pending_clients_notifier = createEventNotifier();
|
||||
if (aeCreateFileEvent(t->el, getReadEventFd(t->pending_clients_notifier),
|
||||
AE_READABLE, handleClientsFromMainThread, t) != AE_OK)
|
||||
{
|
||||
serverLog(LL_WARNING, "Fatal: Can't register file event for IO thread notifications.");
|
||||
exit(1);
|
||||
}
|
||||
|
||||
/* Create IO thread */
|
||||
if (pthread_create(&t->tid, NULL, IOThreadMain, (void*)t) != 0) {
|
||||
serverLog(LL_WARNING, "Fatal: Can't initialize IO thread.");
|
||||
exit(1);
|
||||
}
|
||||
|
||||
/* For main thread */
|
||||
mainThreadPendingClientsToIOThreads[i] = listCreate();
|
||||
mainThreadPendingClients[i] = listCreate();
|
||||
mainThreadProcessingClients[i] = listCreate();
|
||||
pthread_mutex_init(&mainThreadPendingClientsMutexes[i], attr);
|
||||
mainThreadPendingClientsNotifiers[i] = createEventNotifier();
|
||||
if (aeCreateFileEvent(server.el, getReadEventFd(mainThreadPendingClientsNotifiers[i]),
|
||||
AE_READABLE, handleClientsFromIOThread, t) != AE_OK)
|
||||
{
|
||||
serverLog(LL_WARNING, "Fatal: Can't register file event for main thread notifications.");
|
||||
exit(1);
|
||||
}
|
||||
if (attr) zfree(attr);
|
||||
}
|
||||
}
|
||||
|
||||
/* Kill the IO threads, TODO: release the applied resources. */
|
||||
void killIOThreads(void) {
|
||||
if (server.io_threads_num <= 1) return;
|
||||
|
||||
int err, j;
|
||||
for (j = 1; j < server.io_threads_num; j++) {
|
||||
if (IOThreads[j].tid == pthread_self()) continue;
|
||||
if (IOThreads[j].tid && pthread_cancel(IOThreads[j].tid) == 0) {
|
||||
if ((err = pthread_join(IOThreads[j].tid,NULL)) != 0) {
|
||||
serverLog(LL_WARNING,
|
||||
"IO thread(tid:%lu) can not be joined: %s",
|
||||
(unsigned long)IOThreads[j].tid, strerror(err));
|
||||
} else {
|
||||
serverLog(LL_WARNING,
|
||||
"IO thread(tid:%lu) terminated",(unsigned long)IOThreads[j].tid);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
@ -355,7 +355,12 @@ int isWatchedKeyExpired(client *c) {
|
|||
}
|
||||
|
||||
/* "Touch" a key, so that if this key is being WATCHed by some client the
|
||||
* next EXEC will fail. */
|
||||
* next EXEC will fail.
|
||||
*
|
||||
* Sanitizer suppression: IO threads also read c->flags, but never modify
|
||||
* it or read the CLIENT_DIRTY_CAS bit, main thread just only modifies
|
||||
* this bit, so there is actually no real data race. */
|
||||
REDIS_NO_SANITIZE("thread")
|
||||
void touchWatchedKey(redisDb *db, robj *key) {
|
||||
list *clients;
|
||||
listIter li;
|
||||
|
|
@ -404,6 +409,7 @@ void touchWatchedKey(redisDb *db, robj *key) {
|
|||
* replaced_with: for SWAPDB, the WATCH should be invalidated if
|
||||
* the key exists in either of them, and skipped only if it
|
||||
* doesn't exist in both. */
|
||||
REDIS_NO_SANITIZE("thread")
|
||||
void touchAllWatchedKeysInDb(redisDb *emptied, redisDb *replaced_with) {
|
||||
listIter li;
|
||||
listNode *ln;
|
||||
|
|
|
|||
708
src/networking.c
708
src/networking.c
|
|
@ -24,7 +24,6 @@
|
|||
|
||||
static void setProtocolError(const char *errstr, client *c);
|
||||
static void pauseClientsByClient(mstime_t end, int isPauseClientAll);
|
||||
int postponeClientRead(client *c);
|
||||
char *getClientSockname(client *c);
|
||||
static inline int clientTypeIsSlave(client *c);
|
||||
int ProcessingEventsWhileBlocked = 0; /* See processEventsWhileBlocked(). */
|
||||
|
|
@ -132,6 +131,9 @@ client *createClient(connection *conn) {
|
|||
uint64_t client_id;
|
||||
atomicGetIncr(server.next_client_id, client_id, 1);
|
||||
c->id = client_id;
|
||||
c->tid = IOTHREAD_MAIN_THREAD_ID;
|
||||
c->running_tid = IOTHREAD_MAIN_THREAD_ID;
|
||||
if (conn) server.io_threads_clients_num[c->tid]++;
|
||||
#ifdef LOG_REQ_RES
|
||||
reqresReset(c, 0);
|
||||
c->resp = server.client_default_resp;
|
||||
|
|
@ -163,6 +165,8 @@ client *createClient(connection *conn) {
|
|||
c->bulklen = -1;
|
||||
c->sentlen = 0;
|
||||
c->flags = 0;
|
||||
c->io_flags = CLIENT_IO_READ_ENABLED | CLIENT_IO_WRITE_ENABLED;
|
||||
c->read_error = 0;
|
||||
c->slot = -1;
|
||||
c->ctime = c->lastinteraction = server.unixtime;
|
||||
c->duration = 0;
|
||||
|
|
@ -195,8 +199,8 @@ client *createClient(connection *conn) {
|
|||
c->peerid = NULL;
|
||||
c->sockname = NULL;
|
||||
c->client_list_node = NULL;
|
||||
c->io_thread_client_list_node = NULL;
|
||||
c->postponed_list_node = NULL;
|
||||
c->pending_read_list_node = NULL;
|
||||
c->client_tracking_redirection = 0;
|
||||
c->client_tracking_prefixes = NULL;
|
||||
c->last_memory_usage = 0;
|
||||
|
|
@ -300,13 +304,8 @@ int prepareClientToWrite(client *c) {
|
|||
if (!c->conn) return C_ERR; /* Fake client for AOF loading. */
|
||||
|
||||
/* Schedule the client to write the output buffers to the socket, unless
|
||||
* it should already be setup to do so (it has already pending data).
|
||||
*
|
||||
* If CLIENT_PENDING_READ is set, we're in an IO thread and should
|
||||
* not put the client in pending write queue. Instead, it will be
|
||||
* done by handleClientsWithPendingReadsUsingThreads() upon return.
|
||||
*/
|
||||
if (!clientHasPendingReplies(c) && io_threads_op == IO_THREADS_OP_IDLE)
|
||||
* it should already be setup to do so (it has already pending data). */
|
||||
if (!clientHasPendingReplies(c) && likely(c->running_tid == IOTHREAD_MAIN_THREAD_ID))
|
||||
putClientInPendingWriteQueue(c);
|
||||
|
||||
/* Authorize the caller to queue in the output buffer of this client. */
|
||||
|
|
@ -1359,6 +1358,9 @@ void clientAcceptHandler(connection *conn) {
|
|||
moduleFireServerEvent(REDISMODULE_EVENT_CLIENT_CHANGE,
|
||||
REDISMODULE_SUBEVENT_CLIENT_CHANGE_CONNECTED,
|
||||
c);
|
||||
|
||||
/* Assign the client to an IO thread */
|
||||
if (server.io_threads_num > 1) assignClientToIOThread(c);
|
||||
}
|
||||
|
||||
void acceptCommonHandler(connection *conn, int flags, char *ip) {
|
||||
|
|
@ -1547,14 +1549,6 @@ void unlinkClient(client *c) {
|
|||
c->flags &= ~CLIENT_PENDING_WRITE;
|
||||
}
|
||||
|
||||
/* Remove from the list of pending reads if needed. */
|
||||
serverAssert(!c->conn || io_threads_op == IO_THREADS_OP_IDLE);
|
||||
if (c->pending_read_list_node != NULL) {
|
||||
listDelNode(server.clients_pending_read,c->pending_read_list_node);
|
||||
c->pending_read_list_node = NULL;
|
||||
}
|
||||
|
||||
|
||||
/* When client was just unblocked because of a blocking operation,
|
||||
* remove it from the list of unblocked clients. */
|
||||
if (c->flags & CLIENT_UNBLOCKED) {
|
||||
|
|
@ -1631,7 +1625,7 @@ void deauthenticateAndCloseClient(client *c) {
|
|||
* If any data remained in the buffer, the client will take ownership of the buffer
|
||||
* and a new empty buffer will be allocated for the reusable buffer. */
|
||||
static void resetReusableQueryBuf(client *c) {
|
||||
serverAssert(c->flags & CLIENT_REUSABLE_QUERYBUFFER);
|
||||
serverAssert(c->io_flags & CLIENT_IO_REUSABLE_QUERYBUFFER);
|
||||
if (c->querybuf != thread_reusable_qb || sdslen(c->querybuf) > c->qb_pos) {
|
||||
/* If querybuf has been reallocated or there is still data left,
|
||||
* let the client take ownership of the reusable buffer. */
|
||||
|
|
@ -1645,7 +1639,7 @@ static void resetReusableQueryBuf(client *c) {
|
|||
|
||||
/* Mark that the client is no longer using the reusable query buffer
|
||||
* and indicate that it is no longer used by any client. */
|
||||
c->flags &= ~CLIENT_REUSABLE_QUERYBUFFER;
|
||||
c->io_flags &= ~CLIENT_IO_REUSABLE_QUERYBUFFER;
|
||||
thread_reusable_qb_used = 0;
|
||||
}
|
||||
|
||||
|
|
@ -1659,6 +1653,19 @@ void freeClient(client *c) {
|
|||
return;
|
||||
}
|
||||
|
||||
/* If the client is running in io thread, we can't free it directly. */
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
fetchClientFromIOThread(c);
|
||||
}
|
||||
|
||||
/* We need to unbind connection of client from io thread event loop first. */
|
||||
if (c->tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
unbindClientFromIOThreadEventLoop(c);
|
||||
}
|
||||
|
||||
/* Update the number of clients in the IO thread. */
|
||||
if (c->conn) server.io_threads_clients_num[c->tid]--;
|
||||
|
||||
/* For connected clients, call the disconnection event of modules hooks. */
|
||||
if (c->conn) {
|
||||
moduleFireServerEvent(REDISMODULE_EVENT_CLIENT_CHANGE,
|
||||
|
|
@ -1703,7 +1710,7 @@ void freeClient(client *c) {
|
|||
}
|
||||
|
||||
/* Free the query buffer */
|
||||
if (c->flags & CLIENT_REUSABLE_QUERYBUFFER)
|
||||
if (c->io_flags & CLIENT_IO_REUSABLE_QUERYBUFFER)
|
||||
resetReusableQueryBuf(c);
|
||||
sdsfree(c->querybuf);
|
||||
c->querybuf = NULL;
|
||||
|
|
@ -1816,25 +1823,24 @@ void freeClient(client *c) {
|
|||
* a context where calling freeClient() is not possible, because the client
|
||||
* should be valid for the continuation of the flow of the program. */
|
||||
void freeClientAsync(client *c) {
|
||||
/* We need to handle concurrent access to the server.clients_to_close list
|
||||
* only in the freeClientAsync() function, since it's the only function that
|
||||
* may access the list while Redis uses I/O threads. All the other accesses
|
||||
* are in the context of the main thread while the other threads are
|
||||
* idle. */
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
int main_thread = pthread_equal(pthread_self(), server.main_thread_id);
|
||||
/* Make sure the main thread can access IO thread data safely. */
|
||||
if (main_thread) pauseIOThread(c->tid);
|
||||
if (!(c->flags & CLIENT_IO_CLOSE_ASAP)) {
|
||||
c->io_flags |= CLIENT_IO_CLOSE_ASAP;
|
||||
enqueuePendingClientsToMainThread(c, 1);
|
||||
}
|
||||
if (main_thread) resumeIOThread(c->tid);
|
||||
return;
|
||||
}
|
||||
|
||||
if (c->flags & CLIENT_CLOSE_ASAP || c->flags & CLIENT_SCRIPT) return;
|
||||
c->flags |= CLIENT_CLOSE_ASAP;
|
||||
/* Replicas that was marked as CLIENT_CLOSE_ASAP should not keep the
|
||||
* replication backlog from been trimmed. */
|
||||
if (c->flags & CLIENT_SLAVE) freeReplicaReferencedReplBuffer(c);
|
||||
if (server.io_threads_num == 1) {
|
||||
/* no need to bother with locking if there's just one thread (the main thread) */
|
||||
listAddNodeTail(server.clients_to_close,c);
|
||||
return;
|
||||
}
|
||||
static pthread_mutex_t async_free_queue_mutex = PTHREAD_MUTEX_INITIALIZER;
|
||||
pthread_mutex_lock(&async_free_queue_mutex);
|
||||
listAddNodeTail(server.clients_to_close,c);
|
||||
pthread_mutex_unlock(&async_free_queue_mutex);
|
||||
}
|
||||
|
||||
/* Log errors for invalid use and free the client in async way.
|
||||
|
|
@ -1867,7 +1873,7 @@ int beforeNextClient(client *c) {
|
|||
|
||||
/* Skip the client processing if we're in an IO thread, in that case we'll perform
|
||||
this operation later (this function is called again) in the fan-in stage of the threading mechanism */
|
||||
if (io_threads_op != IO_THREADS_OP_IDLE)
|
||||
if (c && c->running_tid != IOTHREAD_MAIN_THREAD_ID)
|
||||
return C_OK;
|
||||
/* Handle async frees */
|
||||
/* Note: this doesn't make the server.clients_to_close list redundant because of
|
||||
|
|
@ -2052,8 +2058,12 @@ int _writeToClient(client *c, ssize_t *nwritten) {
|
|||
* set to 0. So when handler_installed is set to 0 the function must be
|
||||
* thread safe. */
|
||||
int writeToClient(client *c, int handler_installed) {
|
||||
if (!(c->io_flags & CLIENT_IO_WRITE_ENABLED)) return C_OK;
|
||||
/* Update total number of writes on server */
|
||||
atomicIncr(server.stat_total_writes_processed, 1);
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
atomicIncr(server.stat_io_writes_processed, 1);
|
||||
}
|
||||
|
||||
ssize_t nwritten = 0, totwritten = 0;
|
||||
|
||||
|
|
@ -2107,7 +2117,7 @@ int writeToClient(client *c, int handler_installed) {
|
|||
* is always called with handler_installed set to 0 from threads
|
||||
* so we are fine. */
|
||||
if (handler_installed) {
|
||||
serverAssert(io_threads_op == IO_THREADS_OP_IDLE);
|
||||
/* IO Thread also can do that now. */
|
||||
connSetWriteHandler(c->conn, NULL);
|
||||
}
|
||||
|
||||
|
|
@ -2118,10 +2128,10 @@ int writeToClient(client *c, int handler_installed) {
|
|||
}
|
||||
}
|
||||
/* Update client's memory usage after writing.
|
||||
* Since this isn't thread safe we do this conditionally. In case of threaded writes this is done in
|
||||
* handleClientsWithPendingWritesUsingThreads(). */
|
||||
if (io_threads_op == IO_THREADS_OP_IDLE)
|
||||
* Since this isn't thread safe we do this conditionally. */
|
||||
if (c->running_tid == IOTHREAD_MAIN_THREAD_ID) {
|
||||
updateClientMemUsageAndBucket(c);
|
||||
}
|
||||
return C_OK;
|
||||
}
|
||||
|
||||
|
|
@ -2153,6 +2163,15 @@ int handleClientsWithPendingWrites(void) {
|
|||
/* Don't write to clients that are going to be closed anyway. */
|
||||
if (c->flags & CLIENT_CLOSE_ASAP) continue;
|
||||
|
||||
/* Let IO thread handle the client if possible. */
|
||||
if (server.io_threads_num > 1 &&
|
||||
!(c->flags & CLIENT_CLOSE_AFTER_REPLY) &&
|
||||
!isClientMustHandledByMainThread(c))
|
||||
{
|
||||
assignClientToIOThread(c);
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Try to write buffers to the client socket. */
|
||||
if (writeToClient(c,0) == C_ERR) continue;
|
||||
|
||||
|
|
@ -2227,7 +2246,7 @@ void resetClient(client *c) {
|
|||
* path, it is not really released, but only marked for later release. */
|
||||
void protectClient(client *c) {
|
||||
c->flags |= CLIENT_PROTECTED;
|
||||
if (c->conn) {
|
||||
if (c->conn && c->tid == IOTHREAD_MAIN_THREAD_ID) {
|
||||
connSetReadHandler(c->conn,NULL);
|
||||
connSetWriteHandler(c->conn,NULL);
|
||||
}
|
||||
|
|
@ -2238,7 +2257,8 @@ void unprotectClient(client *c) {
|
|||
if (c->flags & CLIENT_PROTECTED) {
|
||||
c->flags &= ~CLIENT_PROTECTED;
|
||||
if (c->conn) {
|
||||
connSetReadHandler(c->conn,readQueryFromClient);
|
||||
if (c->tid == IOTHREAD_MAIN_THREAD_ID)
|
||||
connSetReadHandler(c->conn,readQueryFromClient);
|
||||
if (clientHasPendingReplies(c)) putClientInPendingWriteQueue(c);
|
||||
}
|
||||
}
|
||||
|
|
@ -2263,8 +2283,7 @@ int processInlineBuffer(client *c) {
|
|||
/* Nothing to do without a \r\n */
|
||||
if (newline == NULL) {
|
||||
if (sdslen(c->querybuf)-c->qb_pos > PROTO_INLINE_MAX_SIZE) {
|
||||
addReplyError(c,"Protocol error: too big inline request");
|
||||
setProtocolError("too big inline request",c);
|
||||
c->read_error = CLIENT_READ_TOO_BIG_INLINE_REQUEST;
|
||||
}
|
||||
return C_ERR;
|
||||
}
|
||||
|
|
@ -2279,8 +2298,7 @@ int processInlineBuffer(client *c) {
|
|||
argv = sdssplitargs(aux,&argc);
|
||||
sdsfree(aux);
|
||||
if (argv == NULL) {
|
||||
addReplyError(c,"Protocol error: unbalanced quotes in request");
|
||||
setProtocolError("unbalanced quotes in inline request",c);
|
||||
c->read_error = CLIENT_READ_UNBALANCED_QUOTES;
|
||||
return C_ERR;
|
||||
}
|
||||
|
||||
|
|
@ -2299,8 +2317,7 @@ int processInlineBuffer(client *c) {
|
|||
* to keep the connection active. */
|
||||
if (querylen != 0 && c->flags & CLIENT_MASTER) {
|
||||
sdsfreesplitres(argv,argc);
|
||||
serverLog(LL_WARNING,"WARNING: Receiving inline protocol from master, master stream corruption? Closing the master connection and discarding the cached master.");
|
||||
setProtocolError("Master using the inline protocol. Desync?",c);
|
||||
c->read_error = CLIENT_READ_MASTER_USING_INLINE_PROTOCAL;
|
||||
return C_ERR;
|
||||
}
|
||||
|
||||
|
|
@ -2385,8 +2402,7 @@ int processMultibulkBuffer(client *c) {
|
|||
newline = strchr(c->querybuf+c->qb_pos,'\r');
|
||||
if (newline == NULL) {
|
||||
if (sdslen(c->querybuf)-c->qb_pos > PROTO_INLINE_MAX_SIZE) {
|
||||
addReplyError(c,"Protocol error: too big mbulk count string");
|
||||
setProtocolError("too big mbulk count string",c);
|
||||
c->read_error = CLIENT_READ_TOO_BIG_MBULK_COUNT_STRING;
|
||||
}
|
||||
return C_ERR;
|
||||
}
|
||||
|
|
@ -2400,12 +2416,10 @@ int processMultibulkBuffer(client *c) {
|
|||
serverAssertWithInfo(c,NULL,c->querybuf[c->qb_pos] == '*');
|
||||
ok = string2ll(c->querybuf+1+c->qb_pos,newline-(c->querybuf+1+c->qb_pos),&ll);
|
||||
if (!ok || ll > INT_MAX) {
|
||||
addReplyError(c,"Protocol error: invalid multibulk length");
|
||||
setProtocolError("invalid mbulk count",c);
|
||||
c->read_error = CLIENT_READ_INVALID_MULTIBUCK_LENGTH;
|
||||
return C_ERR;
|
||||
} else if (ll > 10 && authRequired(c)) {
|
||||
addReplyError(c, "Protocol error: unauthenticated multibulk length");
|
||||
setProtocolError("unauth mbulk count", c);
|
||||
c->read_error = CLIENT_READ_UNAUTH_MBUCK_COUNT;
|
||||
return C_ERR;
|
||||
}
|
||||
|
||||
|
|
@ -2432,9 +2446,7 @@ int processMultibulkBuffer(client *c) {
|
|||
newline = strchr(c->querybuf+c->qb_pos,'\r');
|
||||
if (newline == NULL) {
|
||||
if (sdslen(c->querybuf)-c->qb_pos > PROTO_INLINE_MAX_SIZE) {
|
||||
addReplyError(c,
|
||||
"Protocol error: too big bulk count string");
|
||||
setProtocolError("too big bulk count string",c);
|
||||
c->read_error = CLIENT_READ_TOO_BIG_BUCK_COUNT_STRING;
|
||||
return C_ERR;
|
||||
}
|
||||
break;
|
||||
|
|
@ -2445,22 +2457,17 @@ int processMultibulkBuffer(client *c) {
|
|||
break;
|
||||
|
||||
if (c->querybuf[c->qb_pos] != '$') {
|
||||
addReplyErrorFormat(c,
|
||||
"Protocol error: expected '$', got '%c'",
|
||||
c->querybuf[c->qb_pos]);
|
||||
setProtocolError("expected $ but got something else",c);
|
||||
c->read_error = CLIENT_READ_EXPECTED_DOLLAR;
|
||||
return C_ERR;
|
||||
}
|
||||
|
||||
ok = string2ll(c->querybuf+c->qb_pos+1,newline-(c->querybuf+c->qb_pos+1),&ll);
|
||||
if (!ok || ll < 0 ||
|
||||
(!(c->flags & CLIENT_MASTER) && ll > server.proto_max_bulk_len)) {
|
||||
addReplyError(c,"Protocol error: invalid bulk length");
|
||||
setProtocolError("invalid bulk length",c);
|
||||
c->read_error = CLIENT_READ_INVALID_BUCK_LENGTH;
|
||||
return C_ERR;
|
||||
} else if (ll > 16384 && authRequired(c)) {
|
||||
addReplyError(c, "Protocol error: unauthenticated bulk length");
|
||||
setProtocolError("unauth bulk length", c);
|
||||
c->read_error = CLIENT_READ_UNAUTH_BUCK_LENGTH;
|
||||
return C_ERR;
|
||||
}
|
||||
|
||||
|
|
@ -2637,6 +2644,74 @@ int processPendingCommandAndInputBuffer(client *c) {
|
|||
return C_OK;
|
||||
}
|
||||
|
||||
void handleClientReadError(client *c) {
|
||||
switch (c->read_error) {
|
||||
case CLIENT_READ_TOO_BIG_INLINE_REQUEST:
|
||||
addReplyError(c,"Protocol error: too big inline request");
|
||||
setProtocolError("too big inline request",c);
|
||||
break;
|
||||
case CLIENT_READ_UNBALANCED_QUOTES:
|
||||
addReplyError(c,"Protocol error: unbalanced quotes in request");
|
||||
setProtocolError("unbalanced quotes in request",c);
|
||||
break;
|
||||
case CLIENT_READ_MASTER_USING_INLINE_PROTOCAL:
|
||||
serverLog(LL_WARNING,"WARNING: Receiving inline protocol from master, master stream corruption? Closing the master connection and discarding the cached master.");
|
||||
setProtocolError("Master using the inline protocol. Desync?",c);
|
||||
break;
|
||||
case CLIENT_READ_TOO_BIG_MBULK_COUNT_STRING:
|
||||
addReplyError(c,"Protocol error: too big mbulk count string");
|
||||
setProtocolError("too big mbulk count string",c);
|
||||
break;
|
||||
case CLIENT_READ_TOO_BIG_BUCK_COUNT_STRING:
|
||||
addReplyError(c, "Protocol error: too big bulk count string");
|
||||
setProtocolError("too big bulk count string",c);
|
||||
break;
|
||||
case CLIENT_READ_EXPECTED_DOLLAR:
|
||||
addReplyErrorFormat(c,
|
||||
"Protocol error: expected '$', got '%c'",
|
||||
c->querybuf[c->qb_pos]);
|
||||
setProtocolError("expected $ but got something else",c);
|
||||
break;
|
||||
case CLIENT_READ_INVALID_BUCK_LENGTH:
|
||||
addReplyError(c,"Protocol error: invalid bulk length");
|
||||
setProtocolError("invalid bulk length",c);
|
||||
break;
|
||||
case CLIENT_READ_UNAUTH_BUCK_LENGTH:
|
||||
addReplyError(c, "Protocol error: unauthenticated bulk length");
|
||||
setProtocolError("unauth bulk length", c);
|
||||
break;
|
||||
case CLIENT_READ_INVALID_MULTIBUCK_LENGTH:
|
||||
addReplyError(c,"Protocol error: invalid multibulk length");
|
||||
setProtocolError("invalid mbulk count",c);
|
||||
break;
|
||||
case CLIENT_READ_UNAUTH_MBUCK_COUNT:
|
||||
addReplyError(c, "Protocol error: unauthenticated multibulk length");
|
||||
setProtocolError("unauth mbulk count", c);
|
||||
break;
|
||||
case CLIENT_READ_CONN_DISCONNECTED:
|
||||
serverLog(LL_VERBOSE, "Reading from client: %s",connGetLastError(c->conn));
|
||||
break;
|
||||
case CLIENT_READ_CONN_CLOSED:
|
||||
if (server.verbosity <= LL_VERBOSE) {
|
||||
sds info = catClientInfoString(sdsempty(), c);
|
||||
serverLog(LL_VERBOSE, "Client closed connection %s", info);
|
||||
sdsfree(info);
|
||||
}
|
||||
break;
|
||||
case CLIENT_READ_REACHED_MAX_QUERYBUF: {
|
||||
sds ci = catClientInfoString(sdsempty(),c), bytes = sdsempty();
|
||||
bytes = sdscatrepr(bytes,c->querybuf,64);
|
||||
serverLog(LL_WARNING,"Closing client that reached max query buffer length: %s (qbuf initial bytes: %s)", ci, bytes);
|
||||
sdsfree(ci);
|
||||
sdsfree(bytes);
|
||||
break;
|
||||
}
|
||||
default:
|
||||
serverPanic("Unknown client read error");
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
/* This function is called every time, in the client structure 'c', there is
|
||||
* more query buffer to process, because we read more data from the socket
|
||||
* or because a client was blocked and later reactivated, so there could be
|
||||
|
|
@ -2656,7 +2731,7 @@ int processInputBuffer(client *c) {
|
|||
* condition on the slave. We want just to accumulate the replication
|
||||
* stream (instead of replying -BUSY like we do with other clients) and
|
||||
* later resume the processing. */
|
||||
if (isInsideYieldingLongCommand() && c->flags & CLIENT_MASTER) break;
|
||||
if (c->flags & CLIENT_MASTER && isInsideYieldingLongCommand()) break;
|
||||
|
||||
/* CLIENT_CLOSE_AFTER_REPLY closes the connection once the reply is
|
||||
* written to the client. Make sure to not let the reply grow after
|
||||
|
|
@ -2675,23 +2750,34 @@ int processInputBuffer(client *c) {
|
|||
}
|
||||
|
||||
if (c->reqtype == PROTO_REQ_INLINE) {
|
||||
if (processInlineBuffer(c) != C_OK) break;
|
||||
if (processInlineBuffer(c) != C_OK) {
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID && c->read_error)
|
||||
enqueuePendingClientsToMainThread(c, 0);
|
||||
break;
|
||||
}
|
||||
} else if (c->reqtype == PROTO_REQ_MULTIBULK) {
|
||||
if (processMultibulkBuffer(c) != C_OK) break;
|
||||
if (processMultibulkBuffer(c) != C_OK) {
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID && c->read_error)
|
||||
enqueuePendingClientsToMainThread(c, 0);
|
||||
break;
|
||||
}
|
||||
} else {
|
||||
serverPanic("Unknown request type");
|
||||
}
|
||||
|
||||
/* Multibulk processing could see a <= 0 length. */
|
||||
if (c->argc == 0) {
|
||||
resetClientInternal(c, 0);
|
||||
freeClientArgvInternal(c, 0);
|
||||
c->reqtype = 0;
|
||||
c->multibulklen = 0;
|
||||
c->bulklen = -1;
|
||||
} else {
|
||||
/* If we are in the context of an I/O thread, we can't really
|
||||
* execute the command here. All we can do is to flag the client
|
||||
* as one that needs to process the command. */
|
||||
if (io_threads_op != IO_THREADS_OP_IDLE) {
|
||||
serverAssert(io_threads_op == IO_THREADS_OP_READ);
|
||||
c->flags |= CLIENT_PENDING_COMMAND;
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
c->io_flags |= CLIENT_IO_PENDING_COMMAND;
|
||||
enqueuePendingClientsToMainThread(c, 0);
|
||||
break;
|
||||
}
|
||||
|
||||
|
|
@ -2732,7 +2818,7 @@ int processInputBuffer(client *c) {
|
|||
/* Update client memory usage after processing the query buffer, this is
|
||||
* important in case the query buffer is big and wasn't drained during
|
||||
* the above loop (because of partially sent big commands). */
|
||||
if (io_threads_op == IO_THREADS_OP_IDLE)
|
||||
if (c->running_tid == IOTHREAD_MAIN_THREAD_ID)
|
||||
updateClientMemUsageAndBucket(c);
|
||||
|
||||
return C_OK;
|
||||
|
|
@ -2742,13 +2828,14 @@ void readQueryFromClient(connection *conn) {
|
|||
client *c = connGetPrivateData(conn);
|
||||
int nread, big_arg = 0;
|
||||
size_t qblen, readlen;
|
||||
|
||||
/* Check if we want to read from the client later when exiting from
|
||||
* the event loop. This is the case if threaded I/O is enabled. */
|
||||
if (postponeClientRead(c)) return;
|
||||
if (!(c->io_flags & CLIENT_IO_READ_ENABLED)) return;
|
||||
c->read_error = 0;
|
||||
|
||||
/* Update total number of reads on server */
|
||||
atomicIncr(server.stat_total_reads_processed, 1);
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
atomicIncr(server.stat_io_reads_processed, 1);
|
||||
}
|
||||
|
||||
readlen = PROTO_IOBUF_LEN;
|
||||
/* If this is a multi bulk request, and we are processing a bulk reply
|
||||
|
|
@ -2793,7 +2880,7 @@ void readQueryFromClient(connection *conn) {
|
|||
/* Assign the reusable query buffer to the client and mark it as in use. */
|
||||
serverAssert(sdslen(thread_reusable_qb) == 0);
|
||||
c->querybuf = thread_reusable_qb;
|
||||
c->flags |= CLIENT_REUSABLE_QUERYBUFFER;
|
||||
c->io_flags |= CLIENT_IO_REUSABLE_QUERYBUFFER;
|
||||
thread_reusable_qb_used = 1;
|
||||
}
|
||||
}
|
||||
|
|
@ -2821,16 +2908,12 @@ void readQueryFromClient(connection *conn) {
|
|||
if (connGetState(conn) == CONN_STATE_CONNECTED) {
|
||||
goto done;
|
||||
} else {
|
||||
serverLog(LL_VERBOSE, "Reading from client: %s",connGetLastError(c->conn));
|
||||
c->read_error = CLIENT_READ_CONN_DISCONNECTED;
|
||||
freeClientAsync(c);
|
||||
goto done;
|
||||
}
|
||||
} else if (nread == 0) {
|
||||
if (server.verbosity <= LL_VERBOSE) {
|
||||
sds info = catClientInfoString(sdsempty(), c);
|
||||
serverLog(LL_VERBOSE, "Client closed connection %s", info);
|
||||
sdsfree(info);
|
||||
}
|
||||
c->read_error = CLIENT_READ_CONN_CLOSED;
|
||||
freeClientAsync(c);
|
||||
goto done;
|
||||
}
|
||||
|
|
@ -2853,13 +2936,9 @@ void readQueryFromClient(connection *conn) {
|
|||
*
|
||||
* For unauthenticated clients, the query buffer cannot exceed 1MB at most. */
|
||||
(c->mstate.argv_len_sums + sdslen(c->querybuf) > server.client_max_querybuf_len ||
|
||||
(c->mstate.argv_len_sums + sdslen(c->querybuf) > 1024*1024 && authRequired(c)))) {
|
||||
sds ci = catClientInfoString(sdsempty(),c), bytes = sdsempty();
|
||||
|
||||
bytes = sdscatrepr(bytes,c->querybuf,64);
|
||||
serverLog(LL_WARNING,"Closing client that reached max query buffer length: %s (qbuf initial bytes: %s)", ci, bytes);
|
||||
sdsfree(ci);
|
||||
sdsfree(bytes);
|
||||
(c->mstate.argv_len_sums + sdslen(c->querybuf) > 1024*1024 && authRequired(c))))
|
||||
{
|
||||
c->read_error = CLIENT_READ_REACHED_MAX_QUERYBUF;
|
||||
freeClientAsync(c);
|
||||
atomicIncr(server.stat_client_qbuf_limit_disconnections, 1);
|
||||
goto done;
|
||||
|
|
@ -2871,7 +2950,13 @@ void readQueryFromClient(connection *conn) {
|
|||
c = NULL;
|
||||
|
||||
done:
|
||||
if (c && (c->flags & CLIENT_REUSABLE_QUERYBUFFER)) {
|
||||
if (c && c->read_error) {
|
||||
if (c->running_tid == IOTHREAD_MAIN_THREAD_ID) {
|
||||
handleClientReadError(c);
|
||||
}
|
||||
}
|
||||
|
||||
if (c && (c->io_flags & CLIENT_IO_REUSABLE_QUERYBUFFER)) {
|
||||
serverAssert(c->qb_pos == 0); /* Ensure the client's query buffer is trimmed in processInputBuffer */
|
||||
resetReusableQueryBuf(c);
|
||||
}
|
||||
|
|
@ -2933,6 +3018,16 @@ char *getClientSockname(client *c) {
|
|||
sds catClientInfoString(sds s, client *client) {
|
||||
char flags[17], events[3], conninfo[CONN_INFO_LEN], *p;
|
||||
|
||||
/* Pause IO thread to access data of the client safely. */
|
||||
int paused = 0;
|
||||
if (client->running_tid != IOTHREAD_MAIN_THREAD_ID &&
|
||||
pthread_equal(server.main_thread_id, pthread_self()) &&
|
||||
!server.crashing)
|
||||
{
|
||||
paused = 1;
|
||||
pauseIOThread(client->running_tid);
|
||||
}
|
||||
|
||||
p = flags;
|
||||
if (client->flags & CLIENT_SLAVE) {
|
||||
if (client->flags & CLIENT_MONITOR)
|
||||
|
|
@ -3006,7 +3101,10 @@ sds catClientInfoString(sds s, client *client) {
|
|||
" redir=%I", (client->flags & CLIENT_TRACKING) ? (long long) client->client_tracking_redirection : -1,
|
||||
" resp=%i", client->resp,
|
||||
" lib-name=%s", client->lib_name ? (char*)client->lib_name->ptr : "",
|
||||
" lib-ver=%s", client->lib_ver ? (char*)client->lib_ver->ptr : ""));
|
||||
" lib-ver=%s", client->lib_ver ? (char*)client->lib_ver->ptr : "",
|
||||
" io-thread=%i", client->tid));
|
||||
|
||||
if (paused) resumeIOThread(client->running_tid);
|
||||
return ret;
|
||||
}
|
||||
|
||||
|
|
@ -3016,6 +3114,17 @@ sds getAllClientsInfoString(int type) {
|
|||
client *client;
|
||||
sds o = sdsnewlen(SDS_NOINIT,200*listLength(server.clients));
|
||||
sdsclear(o);
|
||||
|
||||
/* Pause all IO threads to access data of clients safely, and pausing the
|
||||
* specific IO thread will not repeatedly execute in catClientInfoString. */
|
||||
int allpaused = 0;
|
||||
if (server.io_threads_num > 1 && !server.crashing &&
|
||||
pthread_equal(server.main_thread_id, pthread_self()))
|
||||
{
|
||||
allpaused = 1;
|
||||
pauseAllIOThreads();
|
||||
}
|
||||
|
||||
listRewind(server.clients,&li);
|
||||
while ((ln = listNext(&li)) != NULL) {
|
||||
client = listNodeValue(ln);
|
||||
|
|
@ -3023,6 +3132,8 @@ sds getAllClientsInfoString(int type) {
|
|||
o = catClientInfoString(o,client);
|
||||
o = sdscatlen(o,"\n",1);
|
||||
}
|
||||
|
||||
if (allpaused) resumeAllIOThreads();
|
||||
return o;
|
||||
}
|
||||
|
||||
|
|
@ -4331,388 +4442,6 @@ void processEventsWhileBlocked(void) {
|
|||
server.cmd_time_snapshot = prev_cmd_time_snapshot;
|
||||
}
|
||||
|
||||
/* ==========================================================================
|
||||
* Threaded I/O
|
||||
* ========================================================================== */
|
||||
|
||||
#define IO_THREADS_MAX_NUM 128
|
||||
|
||||
typedef struct __attribute__((aligned(CACHE_LINE_SIZE))) threads_pending {
|
||||
redisAtomic unsigned long value;
|
||||
} threads_pending;
|
||||
|
||||
pthread_t io_threads[IO_THREADS_MAX_NUM];
|
||||
pthread_mutex_t io_threads_mutex[IO_THREADS_MAX_NUM];
|
||||
threads_pending io_threads_pending[IO_THREADS_MAX_NUM];
|
||||
int io_threads_op; /* IO_THREADS_OP_IDLE, IO_THREADS_OP_READ or IO_THREADS_OP_WRITE. */ // TODO: should access to this be atomic??!
|
||||
|
||||
/* This is the list of clients each thread will serve when threaded I/O is
|
||||
* used. We spawn io_threads_num-1 threads, since one is the main thread
|
||||
* itself. */
|
||||
list *io_threads_list[IO_THREADS_MAX_NUM];
|
||||
|
||||
static inline unsigned long getIOPendingCount(int i) {
|
||||
unsigned long count = 0;
|
||||
atomicGetWithSync(io_threads_pending[i].value, count);
|
||||
return count;
|
||||
}
|
||||
|
||||
static inline void setIOPendingCount(int i, unsigned long count) {
|
||||
atomicSetWithSync(io_threads_pending[i].value, count);
|
||||
}
|
||||
|
||||
void *IOThreadMain(void *myid) {
|
||||
/* The ID is the thread number (from 0 to server.io_threads_num-1), and is
|
||||
* used by the thread to just manipulate a single sub-array of clients. */
|
||||
long id = (unsigned long)myid;
|
||||
char thdname[16];
|
||||
|
||||
snprintf(thdname, sizeof(thdname), "io_thd_%ld", id);
|
||||
redis_set_thread_title(thdname);
|
||||
redisSetCpuAffinity(server.server_cpulist);
|
||||
makeThreadKillable();
|
||||
|
||||
while(1) {
|
||||
/* Wait for start */
|
||||
for (int j = 0; j < 1000000; j++) {
|
||||
if (getIOPendingCount(id) != 0) break;
|
||||
}
|
||||
|
||||
/* Give the main thread a chance to stop this thread. */
|
||||
if (getIOPendingCount(id) == 0) {
|
||||
pthread_mutex_lock(&io_threads_mutex[id]);
|
||||
pthread_mutex_unlock(&io_threads_mutex[id]);
|
||||
continue;
|
||||
}
|
||||
|
||||
serverAssert(getIOPendingCount(id) != 0);
|
||||
|
||||
/* Process: note that the main thread will never touch our list
|
||||
* before we drop the pending count to 0. */
|
||||
listIter li;
|
||||
listNode *ln;
|
||||
listRewind(io_threads_list[id],&li);
|
||||
while((ln = listNext(&li))) {
|
||||
client *c = listNodeValue(ln);
|
||||
if (io_threads_op == IO_THREADS_OP_WRITE) {
|
||||
writeToClient(c,0);
|
||||
} else if (io_threads_op == IO_THREADS_OP_READ) {
|
||||
readQueryFromClient(c->conn);
|
||||
} else {
|
||||
serverPanic("io_threads_op value is unknown");
|
||||
}
|
||||
}
|
||||
listEmpty(io_threads_list[id]);
|
||||
setIOPendingCount(id, 0);
|
||||
}
|
||||
}
|
||||
|
||||
/* Initialize the data structures needed for threaded I/O. */
|
||||
void initThreadedIO(void) {
|
||||
server.io_threads_active = 0; /* We start with threads not active. */
|
||||
|
||||
/* Indicate that io-threads are currently idle */
|
||||
io_threads_op = IO_THREADS_OP_IDLE;
|
||||
|
||||
/* Don't spawn any thread if the user selected a single thread:
|
||||
* we'll handle I/O directly from the main thread. */
|
||||
if (server.io_threads_num == 1) return;
|
||||
|
||||
if (server.io_threads_num > IO_THREADS_MAX_NUM) {
|
||||
serverLog(LL_WARNING,"Fatal: too many I/O threads configured. "
|
||||
"The maximum number is %d.", IO_THREADS_MAX_NUM);
|
||||
exit(1);
|
||||
}
|
||||
|
||||
/* Spawn and initialize the I/O threads. */
|
||||
for (int i = 0; i < server.io_threads_num; i++) {
|
||||
/* Things we do for all the threads including the main thread. */
|
||||
io_threads_list[i] = listCreate();
|
||||
if (i == 0) continue; /* Thread 0 is the main thread. */
|
||||
|
||||
/* Things we do only for the additional threads. */
|
||||
pthread_t tid;
|
||||
pthread_mutex_init(&io_threads_mutex[i],NULL);
|
||||
setIOPendingCount(i, 0);
|
||||
pthread_mutex_lock(&io_threads_mutex[i]); /* Thread will be stopped. */
|
||||
if (pthread_create(&tid,NULL,IOThreadMain,(void*)(long)i) != 0) {
|
||||
serverLog(LL_WARNING,"Fatal: Can't initialize IO thread.");
|
||||
exit(1);
|
||||
}
|
||||
io_threads[i] = tid;
|
||||
}
|
||||
}
|
||||
|
||||
void killIOThreads(void) {
|
||||
int err, j;
|
||||
for (j = 0; j < server.io_threads_num; j++) {
|
||||
if (io_threads[j] == pthread_self()) continue;
|
||||
if (io_threads[j] && pthread_cancel(io_threads[j]) == 0) {
|
||||
if ((err = pthread_join(io_threads[j],NULL)) != 0) {
|
||||
serverLog(LL_WARNING,
|
||||
"IO thread(tid:%lu) can not be joined: %s",
|
||||
(unsigned long)io_threads[j], strerror(err));
|
||||
} else {
|
||||
serverLog(LL_WARNING,
|
||||
"IO thread(tid:%lu) terminated",(unsigned long)io_threads[j]);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void startThreadedIO(void) {
|
||||
serverAssert(server.io_threads_active == 0);
|
||||
for (int j = 1; j < server.io_threads_num; j++)
|
||||
pthread_mutex_unlock(&io_threads_mutex[j]);
|
||||
server.io_threads_active = 1;
|
||||
}
|
||||
|
||||
void stopThreadedIO(void) {
|
||||
/* We may have still clients with pending reads when this function
|
||||
* is called: handle them before stopping the threads. */
|
||||
handleClientsWithPendingReadsUsingThreads();
|
||||
serverAssert(server.io_threads_active == 1);
|
||||
for (int j = 1; j < server.io_threads_num; j++)
|
||||
pthread_mutex_lock(&io_threads_mutex[j]);
|
||||
server.io_threads_active = 0;
|
||||
}
|
||||
|
||||
/* This function checks if there are not enough pending clients to justify
|
||||
* taking the I/O threads active: in that case I/O threads are stopped if
|
||||
* currently active. We track the pending writes as a measure of clients
|
||||
* we need to handle in parallel, however the I/O threading is disabled
|
||||
* globally for reads as well if we have too little pending clients.
|
||||
*
|
||||
* The function returns 0 if the I/O threading should be used because there
|
||||
* are enough active threads, otherwise 1 is returned and the I/O threads
|
||||
* could be possibly stopped (if already active) as a side effect. */
|
||||
int stopThreadedIOIfNeeded(void) {
|
||||
int pending = listLength(server.clients_pending_write);
|
||||
|
||||
/* Return ASAP if IO threads are disabled (single threaded mode). */
|
||||
if (server.io_threads_num == 1) return 1;
|
||||
|
||||
if (pending < (server.io_threads_num*2)) {
|
||||
if (server.io_threads_active) stopThreadedIO();
|
||||
return 1;
|
||||
} else {
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
/* This function achieves thread safety using a fan-out -> fan-in paradigm:
|
||||
* Fan out: The main thread fans out work to the io-threads which block until
|
||||
* setIOPendingCount() is called with a value larger than 0 by the main thread.
|
||||
* Fan in: The main thread waits until getIOPendingCount() returns 0. Then
|
||||
* it can safely perform post-processing and return to normal synchronous
|
||||
* work. */
|
||||
int handleClientsWithPendingWritesUsingThreads(void) {
|
||||
int processed = listLength(server.clients_pending_write);
|
||||
if (processed == 0) return 0; /* Return ASAP if there are no clients. */
|
||||
|
||||
/* If I/O threads are disabled or we have few clients to serve, don't
|
||||
* use I/O threads, but the boring synchronous code. */
|
||||
if (server.io_threads_num == 1 || stopThreadedIOIfNeeded()) {
|
||||
return handleClientsWithPendingWrites();
|
||||
}
|
||||
|
||||
/* Start threads if needed. */
|
||||
if (!server.io_threads_active) startThreadedIO();
|
||||
|
||||
/* Distribute the clients across N different lists. */
|
||||
listIter li;
|
||||
listNode *ln;
|
||||
listRewind(server.clients_pending_write,&li);
|
||||
int item_id = 0;
|
||||
while((ln = listNext(&li))) {
|
||||
client *c = listNodeValue(ln);
|
||||
c->flags &= ~CLIENT_PENDING_WRITE;
|
||||
|
||||
/* Remove clients from the list of pending writes since
|
||||
* they are going to be closed ASAP. */
|
||||
if (c->flags & CLIENT_CLOSE_ASAP) {
|
||||
listUnlinkNode(server.clients_pending_write, ln);
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Since all replicas and replication backlog use global replication
|
||||
* buffer, to guarantee data accessing thread safe, we must put all
|
||||
* replicas client into io_threads_list[0] i.e. main thread handles
|
||||
* sending the output buffer of all replicas. */
|
||||
if (unlikely(clientTypeIsSlave(c))) {
|
||||
listAddNodeTail(io_threads_list[0],c);
|
||||
continue;
|
||||
}
|
||||
|
||||
int target_id = item_id % server.io_threads_num;
|
||||
listAddNodeTail(io_threads_list[target_id],c);
|
||||
item_id++;
|
||||
}
|
||||
|
||||
/* Give the start condition to the waiting threads, by setting the
|
||||
* start condition atomic var. */
|
||||
io_threads_op = IO_THREADS_OP_WRITE;
|
||||
for (int j = 1; j < server.io_threads_num; j++) {
|
||||
int count = listLength(io_threads_list[j]);
|
||||
setIOPendingCount(j, count);
|
||||
}
|
||||
|
||||
/* Also use the main thread to process a slice of clients. */
|
||||
listRewind(io_threads_list[0],&li);
|
||||
while((ln = listNext(&li))) {
|
||||
client *c = listNodeValue(ln);
|
||||
writeToClient(c,0);
|
||||
}
|
||||
listEmpty(io_threads_list[0]);
|
||||
|
||||
/* Wait for all the other threads to end their work. */
|
||||
while(1) {
|
||||
unsigned long pending = 0;
|
||||
for (int j = 1; j < server.io_threads_num; j++)
|
||||
pending += getIOPendingCount(j);
|
||||
if (pending == 0) break;
|
||||
}
|
||||
|
||||
io_threads_op = IO_THREADS_OP_IDLE;
|
||||
|
||||
/* Run the list of clients again to install the write handler where
|
||||
* needed. */
|
||||
listRewind(server.clients_pending_write,&li);
|
||||
while((ln = listNext(&li))) {
|
||||
client *c = listNodeValue(ln);
|
||||
|
||||
/* Update the client in the mem usage after we're done processing it in the io-threads */
|
||||
updateClientMemUsageAndBucket(c);
|
||||
|
||||
/* Install the write handler if there are pending writes in some
|
||||
* of the clients. */
|
||||
if (clientHasPendingReplies(c)) {
|
||||
installClientWriteHandler(c);
|
||||
}
|
||||
}
|
||||
while(listLength(server.clients_pending_write) > 0) {
|
||||
listUnlinkNode(server.clients_pending_write, server.clients_pending_write->head);
|
||||
}
|
||||
|
||||
/* Update processed count on server */
|
||||
server.stat_io_writes_processed += processed;
|
||||
|
||||
return processed;
|
||||
}
|
||||
|
||||
/* Return 1 if we want to handle the client read later using threaded I/O.
|
||||
* This is called by the readable handler of the event loop.
|
||||
* As a side effect of calling this function the client is put in the
|
||||
* pending read clients and flagged as such. */
|
||||
int postponeClientRead(client *c) {
|
||||
if (server.io_threads_active &&
|
||||
server.io_threads_do_reads &&
|
||||
!ProcessingEventsWhileBlocked &&
|
||||
!(c->flags & (CLIENT_MASTER|CLIENT_SLAVE|CLIENT_BLOCKED)) &&
|
||||
io_threads_op == IO_THREADS_OP_IDLE)
|
||||
{
|
||||
listAddNodeHead(server.clients_pending_read,c);
|
||||
c->pending_read_list_node = listFirst(server.clients_pending_read);
|
||||
return 1;
|
||||
} else {
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
/* When threaded I/O is also enabled for the reading + parsing side, the
|
||||
* readable handler will just put normal clients into a queue of clients to
|
||||
* process (instead of serving them synchronously). This function runs
|
||||
* the queue using the I/O threads, and process them in order to accumulate
|
||||
* the reads in the buffers, and also parse the first command available
|
||||
* rendering it in the client structures.
|
||||
* This function achieves thread safety using a fan-out -> fan-in paradigm:
|
||||
* Fan out: The main thread fans out work to the io-threads which block until
|
||||
* setIOPendingCount() is called with a value larger than 0 by the main thread.
|
||||
* Fan in: The main thread waits until getIOPendingCount() returns 0. Then
|
||||
* it can safely perform post-processing and return to normal synchronous
|
||||
* work. */
|
||||
int handleClientsWithPendingReadsUsingThreads(void) {
|
||||
if (!server.io_threads_active || !server.io_threads_do_reads) return 0;
|
||||
int processed = listLength(server.clients_pending_read);
|
||||
if (processed == 0) return 0;
|
||||
|
||||
/* Distribute the clients across N different lists. */
|
||||
listIter li;
|
||||
listNode *ln;
|
||||
listRewind(server.clients_pending_read,&li);
|
||||
int item_id = 0;
|
||||
while((ln = listNext(&li))) {
|
||||
client *c = listNodeValue(ln);
|
||||
int target_id = item_id % server.io_threads_num;
|
||||
listAddNodeTail(io_threads_list[target_id],c);
|
||||
item_id++;
|
||||
}
|
||||
|
||||
/* Give the start condition to the waiting threads, by setting the
|
||||
* start condition atomic var. */
|
||||
io_threads_op = IO_THREADS_OP_READ;
|
||||
for (int j = 1; j < server.io_threads_num; j++) {
|
||||
int count = listLength(io_threads_list[j]);
|
||||
setIOPendingCount(j, count);
|
||||
}
|
||||
|
||||
/* Also use the main thread to process a slice of clients. */
|
||||
listRewind(io_threads_list[0],&li);
|
||||
while((ln = listNext(&li))) {
|
||||
client *c = listNodeValue(ln);
|
||||
readQueryFromClient(c->conn);
|
||||
}
|
||||
listEmpty(io_threads_list[0]);
|
||||
|
||||
/* Wait for all the other threads to end their work. */
|
||||
while(1) {
|
||||
unsigned long pending = 0;
|
||||
for (int j = 1; j < server.io_threads_num; j++)
|
||||
pending += getIOPendingCount(j);
|
||||
if (pending == 0) break;
|
||||
}
|
||||
|
||||
io_threads_op = IO_THREADS_OP_IDLE;
|
||||
|
||||
/* Run the list of clients again to process the new buffers. */
|
||||
while(listLength(server.clients_pending_read)) {
|
||||
ln = listFirst(server.clients_pending_read);
|
||||
client *c = listNodeValue(ln);
|
||||
listDelNode(server.clients_pending_read,ln);
|
||||
c->pending_read_list_node = NULL;
|
||||
|
||||
serverAssert(!(c->flags & CLIENT_BLOCKED));
|
||||
|
||||
if (beforeNextClient(c) == C_ERR) {
|
||||
/* If the client is no longer valid, we avoid
|
||||
* processing the client later. So we just go
|
||||
* to the next. */
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Once io-threads are idle we can update the client in the mem usage */
|
||||
updateClientMemUsageAndBucket(c);
|
||||
|
||||
if (processPendingCommandAndInputBuffer(c) == C_ERR) {
|
||||
/* If the client is no longer valid, we avoid
|
||||
* processing the client later. So we just go
|
||||
* to the next. */
|
||||
continue;
|
||||
}
|
||||
|
||||
/* We may have pending replies if a thread readQueryFromClient() produced
|
||||
* replies and did not put the client in pending write queue (it can't).
|
||||
*/
|
||||
if (!(c->flags & CLIENT_PENDING_WRITE) && clientHasPendingReplies(c))
|
||||
putClientInPendingWriteQueue(c);
|
||||
}
|
||||
|
||||
/* Update processed count on server */
|
||||
server.stat_io_reads_processed += processed;
|
||||
|
||||
return processed;
|
||||
}
|
||||
|
||||
/* Returns the actual client eviction limit based on current configuration or
|
||||
* 0 if no limit. */
|
||||
size_t getClientEvictionLimit(void) {
|
||||
|
|
@ -4752,11 +4481,34 @@ void evictClients(void) {
|
|||
listNode *ln = listNext(&bucket_iter);
|
||||
if (ln) {
|
||||
client *c = ln->value;
|
||||
sds ci = catClientInfoString(sdsempty(),c);
|
||||
serverLog(LL_NOTICE, "Evicting client: %s", ci);
|
||||
freeClient(c);
|
||||
sdsfree(ci);
|
||||
server.stat_evictedclients++;
|
||||
size_t last_memory = c->last_memory_usage;
|
||||
int tid = c->running_tid;
|
||||
if (tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
pauseIOThread(tid);
|
||||
/* We need to update the client memory usage and bucket if the client
|
||||
* is running in IO thread. This is because the client memory usage
|
||||
* and bucket are updated 'only' in the main thread, such as processing
|
||||
* command and clientsCron, it may delay updating, to avoid incorrectly
|
||||
* evicting clients, we update again before evicting, if the memory
|
||||
* used by the client does not decrease or memory usage bucket is not
|
||||
* changed, then we will evict it, otherwise, not evict it. */
|
||||
updateClientMemUsageAndBucket(c);
|
||||
}
|
||||
if (c->last_memory_usage >= last_memory ||
|
||||
c->mem_usage_bucket == &server.client_mem_usage_buckets[curr_bucket])
|
||||
{
|
||||
sds ci = catClientInfoString(sdsempty(),c);
|
||||
serverLog(LL_NOTICE, "Evicting client: %s", ci);
|
||||
freeClient(c);
|
||||
sdsfree(ci);
|
||||
server.stat_evictedclients++;
|
||||
}
|
||||
if (tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
resumeIOThread(tid);
|
||||
/* The 'next' of 'bucket_iter' may be changed after updating client memory
|
||||
* usage and freeing client, so let reset 'bucket_iter'. */
|
||||
listRewind(server.client_mem_usage_buckets[curr_bucket].clients, &bucket_iter);
|
||||
}
|
||||
} else {
|
||||
curr_bucket--;
|
||||
if (curr_bucket < 0) {
|
||||
|
|
|
|||
|
|
@ -2925,7 +2925,7 @@ write_error: /* Handle sendCommand() errors. */
|
|||
}
|
||||
|
||||
int connectWithMaster(void) {
|
||||
server.repl_transfer_s = connCreate(connTypeOfReplication());
|
||||
server.repl_transfer_s = connCreate(server.el, connTypeOfReplication());
|
||||
if (connConnect(server.repl_transfer_s, server.masterhost, server.masterport,
|
||||
server.bind_source_addr, syncWithMaster) == C_ERR) {
|
||||
serverLog(LL_WARNING,"Unable to connect to MASTER: %s",
|
||||
|
|
|
|||
83
src/server.c
83
src/server.c
|
|
@ -963,7 +963,7 @@ void removeClientFromMemUsageBucket(client *c, int allow_eviction) {
|
|||
* returns 1 if client eviction for this client is allowed, 0 otherwise.
|
||||
*/
|
||||
int updateClientMemUsageAndBucket(client *c) {
|
||||
serverAssert(io_threads_op == IO_THREADS_OP_IDLE && c->conn);
|
||||
serverAssert(pthread_equal(pthread_self(), server.main_thread_id) && c->conn);
|
||||
int allow_eviction = clientEvictionAllowed(c);
|
||||
removeClientFromMemUsageBucket(c, allow_eviction);
|
||||
|
||||
|
|
@ -1015,6 +1015,7 @@ void getExpansiveClientsInfo(size_t *in_usage, size_t *out_usage) {
|
|||
* default server.hz value is 10, so sometimes here we need to process thousands
|
||||
* of clients per second, turning this function into a source of latency.
|
||||
*/
|
||||
#define CLIENTS_CRON_PAUSE_IOTHREAD 8
|
||||
#define CLIENTS_CRON_MIN_ITERATIONS 5
|
||||
void clientsCron(void) {
|
||||
/* Try to process at least numclients/server.hz of clients
|
||||
|
|
@ -1049,6 +1050,15 @@ void clientsCron(void) {
|
|||
ClientsPeakMemInput[zeroidx] = 0;
|
||||
ClientsPeakMemOutput[zeroidx] = 0;
|
||||
|
||||
/* Pause the IO threads that are processing clients, to let us access clients
|
||||
* safely. In order to avoid increasing CPU usage by pausing all threads when
|
||||
* there are too many io threads, we pause io threads in multiple batches. */
|
||||
static int start = 1, end = 0;
|
||||
if (server.io_threads_num >= 1 && listLength(server.clients) > 0) {
|
||||
end = start + CLIENTS_CRON_PAUSE_IOTHREAD - 1;
|
||||
if (end >= server.io_threads_num) end = server.io_threads_num - 1;
|
||||
pauseIOThreadsRange(start, end);
|
||||
}
|
||||
|
||||
while(listLength(server.clients) && iterations--) {
|
||||
client *c;
|
||||
|
|
@ -1059,6 +1069,15 @@ void clientsCron(void) {
|
|||
head = listFirst(server.clients);
|
||||
c = listNodeValue(head);
|
||||
listRotateHeadToTail(server.clients);
|
||||
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID &&
|
||||
!(c->running_tid >= start && c->running_tid <= end))
|
||||
{
|
||||
/* Skip clients that are being processed by the IO threads that
|
||||
* are not paused. */
|
||||
continue;
|
||||
}
|
||||
|
||||
/* The following functions do different service checks on the client.
|
||||
* The protocol is that they return non-zero if the client was
|
||||
* terminated. */
|
||||
|
|
@ -1080,6 +1099,14 @@ void clientsCron(void) {
|
|||
|
||||
if (closeClientOnOutputBufferLimitReached(c, 0)) continue;
|
||||
}
|
||||
|
||||
/* Resume the IO threads that were paused */
|
||||
if (end) {
|
||||
resumeIOThreadsRange(start, end);
|
||||
start = end + 1;
|
||||
if (start >= server.io_threads_num) start = 1;
|
||||
end = 0;
|
||||
}
|
||||
}
|
||||
|
||||
/* This function handles 'background' operations we are required to do
|
||||
|
|
@ -1528,9 +1555,6 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
|
|||
migrateCloseTimedoutSockets();
|
||||
}
|
||||
|
||||
/* Stop the I/O threads if we don't have enough pending work. */
|
||||
stopThreadedIOIfNeeded();
|
||||
|
||||
/* Resize tracking keys table if needed. This is also done at every
|
||||
* command execution, but we want to be sure that if the last command
|
||||
* executed changes the value via CONFIG SET, the server will perform
|
||||
|
|
@ -1682,24 +1706,28 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
|
|||
* events to handle. */
|
||||
if (ProcessingEventsWhileBlocked) {
|
||||
uint64_t processed = 0;
|
||||
processed += handleClientsWithPendingReadsUsingThreads();
|
||||
processed += connTypeProcessPendingData();
|
||||
processed += connTypeProcessPendingData(server.el);
|
||||
if (server.aof_state == AOF_ON || server.aof_state == AOF_WAIT_REWRITE)
|
||||
flushAppendOnlyFile(0);
|
||||
processed += handleClientsWithPendingWrites();
|
||||
processed += freeClientsInAsyncFreeQueue();
|
||||
|
||||
/* Let the clients after the blocking call be processed. */
|
||||
processClientsOfAllIOThreads();
|
||||
/* New connections may have been established while blocked, clients from
|
||||
* IO thread may have replies to write, ensure they are promptly sent to
|
||||
* IO threads. */
|
||||
processed += sendPendingClientsToIOThreads();
|
||||
|
||||
server.events_processed_while_blocked += processed;
|
||||
return;
|
||||
}
|
||||
|
||||
/* We should handle pending reads clients ASAP after event loop. */
|
||||
handleClientsWithPendingReadsUsingThreads();
|
||||
|
||||
/* Handle pending data(typical TLS). (must be done before flushAppendOnlyFile) */
|
||||
connTypeProcessPendingData();
|
||||
connTypeProcessPendingData(server.el);
|
||||
|
||||
/* If any connection type(typical TLS) still has pending unread data don't sleep at all. */
|
||||
int dont_sleep = connTypeHasPendingData();
|
||||
int dont_sleep = connTypeHasPendingData(server.el);
|
||||
|
||||
/* Call the Redis Cluster before sleep function. Note that this function
|
||||
* may change the state of Redis Cluster (from ok to fail or vice versa),
|
||||
|
|
@ -1765,8 +1793,8 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
|
|||
long long prev_fsynced_reploff = server.fsynced_reploff;
|
||||
|
||||
/* Write the AOF buffer on disk,
|
||||
* must be done before handleClientsWithPendingWritesUsingThreads,
|
||||
* in case of appendfsync=always. */
|
||||
* must be done before handleClientsWithPendingWrites and
|
||||
* sendPendingClientsToIOThreads, in case of appendfsync=always. */
|
||||
if (server.aof_state == AOF_ON || server.aof_state == AOF_WAIT_REWRITE)
|
||||
flushAppendOnlyFile(0);
|
||||
|
||||
|
|
@ -1788,7 +1816,10 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
|
|||
}
|
||||
|
||||
/* Handle writes with pending output buffers. */
|
||||
handleClientsWithPendingWritesUsingThreads();
|
||||
handleClientsWithPendingWrites();
|
||||
|
||||
/* Let io thread to handle its pending clients. */
|
||||
sendPendingClientsToIOThreads();
|
||||
|
||||
/* Record cron time in beforeSleep. This does not include the time consumed by AOF writing and IO writing above. */
|
||||
monotime cron_start_time_after_write = getMonotonicUs();
|
||||
|
|
@ -2117,6 +2148,7 @@ void initServerConfig(void) {
|
|||
memset(server.blocked_clients_by_type,0,
|
||||
sizeof(server.blocked_clients_by_type));
|
||||
server.shutdown_asap = 0;
|
||||
server.crashing = 0;
|
||||
server.shutdown_flags = 0;
|
||||
server.shutdown_mstime = 0;
|
||||
server.cluster_module_flags = CLUSTER_MODULE_FLAG_NONE;
|
||||
|
|
@ -2583,9 +2615,9 @@ void resetServerStats(void) {
|
|||
server.stat_sync_full = 0;
|
||||
server.stat_sync_partial_ok = 0;
|
||||
server.stat_sync_partial_err = 0;
|
||||
server.stat_io_reads_processed = 0;
|
||||
atomicSet(server.stat_io_reads_processed, 0);
|
||||
atomicSet(server.stat_total_reads_processed, 0);
|
||||
server.stat_io_writes_processed = 0;
|
||||
atomicSet(server.stat_io_writes_processed, 0);
|
||||
atomicSet(server.stat_total_writes_processed, 0);
|
||||
atomicSet(server.stat_client_qbuf_limit_disconnections, 0);
|
||||
server.stat_client_outbuf_limit_disconnections = 0;
|
||||
|
|
@ -2778,6 +2810,7 @@ void initServer(void) {
|
|||
server.aof_last_write_errno = 0;
|
||||
server.repl_good_slaves_count = 0;
|
||||
server.last_sig_received = 0;
|
||||
memset(server.io_threads_clients_num, 0, sizeof(server.io_threads_clients_num));
|
||||
|
||||
/* Initiate acl info struct */
|
||||
server.acl_info.invalid_cmd_accesses = 0;
|
||||
|
|
@ -5535,7 +5568,7 @@ void releaseInfoSectionDict(dict *sec) {
|
|||
* The resulting dictionary should be released with releaseInfoSectionDict. */
|
||||
dict *genInfoSectionDict(robj **argv, int argc, char **defaults, int *out_all, int *out_everything) {
|
||||
char *default_sections[] = {
|
||||
"server", "clients", "memory", "persistence", "stats", "replication",
|
||||
"server", "clients", "memory", "persistence", "stats", "replication", "threads",
|
||||
"cpu", "module_list", "errorstats", "cluster", "keyspace", "keysizes", NULL};
|
||||
if (!defaults)
|
||||
defaults = default_sections;
|
||||
|
|
@ -5886,6 +5919,7 @@ sds genRedisInfoString(dict *section_dict, int all_sections, int everything) {
|
|||
long long current_active_defrag_time = server.stat_last_active_defrag_time ?
|
||||
(long long) elapsedUs(server.stat_last_active_defrag_time): 0;
|
||||
long long stat_client_qbuf_limit_disconnections;
|
||||
long long stat_io_reads_processed, stat_io_writes_processed;
|
||||
atomicGet(server.stat_total_reads_processed, stat_total_reads_processed);
|
||||
atomicGet(server.stat_total_writes_processed, stat_total_writes_processed);
|
||||
atomicGet(server.stat_net_input_bytes, stat_net_input_bytes);
|
||||
|
|
@ -5893,6 +5927,8 @@ sds genRedisInfoString(dict *section_dict, int all_sections, int everything) {
|
|||
atomicGet(server.stat_net_repl_input_bytes, stat_net_repl_input_bytes);
|
||||
atomicGet(server.stat_net_repl_output_bytes, stat_net_repl_output_bytes);
|
||||
atomicGet(server.stat_client_qbuf_limit_disconnections, stat_client_qbuf_limit_disconnections);
|
||||
atomicGet(server.stat_io_reads_processed, stat_io_reads_processed);
|
||||
atomicGet(server.stat_io_writes_processed, stat_io_writes_processed);
|
||||
|
||||
if (sections++) info = sdscat(info,"\r\n");
|
||||
info = sdscatprintf(info, "# Stats\r\n" FMTARGS(
|
||||
|
|
@ -5944,8 +5980,8 @@ sds genRedisInfoString(dict *section_dict, int all_sections, int everything) {
|
|||
"dump_payload_sanitizations:%lld\r\n", server.stat_dump_payload_sanitizations,
|
||||
"total_reads_processed:%lld\r\n", stat_total_reads_processed,
|
||||
"total_writes_processed:%lld\r\n", stat_total_writes_processed,
|
||||
"io_threaded_reads_processed:%lld\r\n", server.stat_io_reads_processed,
|
||||
"io_threaded_writes_processed:%lld\r\n", server.stat_io_writes_processed,
|
||||
"io_threaded_reads_processed:%lld\r\n", stat_io_reads_processed,
|
||||
"io_threaded_writes_processed:%lld\r\n", stat_io_writes_processed,
|
||||
"client_query_buffer_limit_disconnections:%lld\r\n", stat_client_qbuf_limit_disconnections,
|
||||
"client_output_buffer_limit_disconnections:%lld\r\n", server.stat_client_outbuf_limit_disconnections,
|
||||
"reply_buffer_shrinks:%lld\r\n", server.stat_reply_buffer_shrinks,
|
||||
|
|
@ -6094,6 +6130,15 @@ sds genRedisInfoString(dict *section_dict, int all_sections, int everything) {
|
|||
#endif /* RUSAGE_THREAD */
|
||||
}
|
||||
|
||||
/* Threads */
|
||||
if (all_sections || (dictFind(section_dict,"threads") != NULL)) {
|
||||
if (sections++) info = sdscat(info,"\r\n");
|
||||
info = sdscatprintf(info, "# Threads\r\n");
|
||||
for (j = 0; j < server.io_threads_num; j++) {
|
||||
info = sdscatprintf(info, "io_thread_%d:clients=%d\r\n", j, server.io_threads_clients_num[j]);
|
||||
}
|
||||
}
|
||||
|
||||
/* Modules */
|
||||
if (all_sections || (dictFind(section_dict,"module_list") != NULL) || (dictFind(section_dict,"modules") != NULL)) {
|
||||
if (sections++) info = sdscat(info,"\r\n");
|
||||
|
|
|
|||
95
src/server.h
95
src/server.h
|
|
@ -61,6 +61,7 @@ typedef long long ustime_t; /* microsecond time type. */
|
|||
N-elements flat arrays */
|
||||
#include "rax.h" /* Radix tree */
|
||||
#include "connection.h" /* Connection abstraction */
|
||||
#include "eventnotifier.h" /* Event notification */
|
||||
|
||||
#define REDISMODULE_CORE 1
|
||||
typedef struct redisObject robj;
|
||||
|
|
@ -184,6 +185,14 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
|
|||
/* Hash table parameters */
|
||||
#define HASHTABLE_MAX_LOAD_FACTOR 1.618 /* Maximum hash table load factor. */
|
||||
|
||||
/* Max number of IO threads */
|
||||
#define IO_THREADS_MAX_NUM 128
|
||||
|
||||
/* Main thread id for doing IO work, whatever we enable or disable io thread
|
||||
* the main thread always does IO work, so we can consider that the main thread
|
||||
* is the io thread 0. */
|
||||
#define IOTHREAD_MAIN_THREAD_ID 0
|
||||
|
||||
/* Command flags. Please check the definition of struct redisCommand in this file
|
||||
* for more information about the meaning of every flag. */
|
||||
#define CMD_WRITE (1ULL<<0)
|
||||
|
|
@ -385,11 +394,33 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
|
|||
#define CLIENT_MODULE_PREVENT_AOF_PROP (1ULL<<48) /* Module client do not want to propagate to AOF */
|
||||
#define CLIENT_MODULE_PREVENT_REPL_PROP (1ULL<<49) /* Module client do not want to propagate to replica */
|
||||
#define CLIENT_REPROCESSING_COMMAND (1ULL<<50) /* The client is re-processing the command. */
|
||||
#define CLIENT_REUSABLE_QUERYBUFFER (1ULL<<51) /* The client is using the reusable query buffer. */
|
||||
|
||||
/* Any flag that does not let optimize FLUSH SYNC to run it in bg as blocking client ASYNC */
|
||||
#define CLIENT_AVOID_BLOCKING_ASYNC_FLUSH (CLIENT_DENY_BLOCKING|CLIENT_MULTI|CLIENT_LUA_DEBUG|CLIENT_LUA_DEBUG_SYNC|CLIENT_MODULE)
|
||||
|
||||
/* Client flags for client IO */
|
||||
#define CLIENT_IO_READ_ENABLED (1ULL<<0) /* Client can read from socket. */
|
||||
#define CLIENT_IO_WRITE_ENABLED (1ULL<<1) /* Client can write to socket. */
|
||||
#define CLIENT_IO_PENDING_COMMAND (1ULL<<2) /* Similar to CLIENT_PENDING_COMMAND. */
|
||||
#define CLIENT_IO_REUSABLE_QUERYBUFFER (1ULL<<3) /* The client is using the reusable query buffer. */
|
||||
#define CLIENT_IO_CLOSE_ASAP (1ULL<<4) /* Close this client ASAP in IO thread. */
|
||||
|
||||
/* Definitions for client read errors. These error codes are used to indicate
|
||||
* various issues that can occur while reading or parsing data from a client. */
|
||||
#define CLIENT_READ_TOO_BIG_INLINE_REQUEST 1
|
||||
#define CLIENT_READ_UNBALANCED_QUOTES 2
|
||||
#define CLIENT_READ_MASTER_USING_INLINE_PROTOCAL 3
|
||||
#define CLIENT_READ_TOO_BIG_MBULK_COUNT_STRING 4
|
||||
#define CLIENT_READ_TOO_BIG_BUCK_COUNT_STRING 5
|
||||
#define CLIENT_READ_EXPECTED_DOLLAR 6
|
||||
#define CLIENT_READ_INVALID_BUCK_LENGTH 7
|
||||
#define CLIENT_READ_UNAUTH_BUCK_LENGTH 8
|
||||
#define CLIENT_READ_INVALID_MULTIBUCK_LENGTH 9
|
||||
#define CLIENT_READ_UNAUTH_MBUCK_COUNT 10
|
||||
#define CLIENT_READ_CONN_DISCONNECTED 11
|
||||
#define CLIENT_READ_CONN_CLOSED 12
|
||||
#define CLIENT_READ_REACHED_MAX_QUERYBUF 13
|
||||
|
||||
/* Client block type (btype field in client structure)
|
||||
* if CLIENT_BLOCKED flag is set. */
|
||||
typedef enum blocking_type {
|
||||
|
|
@ -578,6 +609,12 @@ typedef enum {
|
|||
#define SHUTDOWN_NOW 4 /* Don't wait for replicas to catch up. */
|
||||
#define SHUTDOWN_FORCE 8 /* Don't let errors prevent shutdown. */
|
||||
|
||||
/* IO thread pause status */
|
||||
#define IO_THREAD_UNPAUSED 0
|
||||
#define IO_THREAD_PAUSING 1
|
||||
#define IO_THREAD_PAUSED 2
|
||||
#define IO_THREAD_RESUMING 3
|
||||
|
||||
/* Command call flags, see call() function */
|
||||
#define CMD_CALL_NONE 0
|
||||
#define CMD_CALL_PROPAGATE_AOF (1<<0)
|
||||
|
|
@ -1159,6 +1196,10 @@ typedef struct client {
|
|||
uint64_t id; /* Client incremental unique ID. */
|
||||
uint64_t flags; /* Client flags: CLIENT_* macros. */
|
||||
connection *conn;
|
||||
uint8_t tid; /* Thread assigned ID this client is bound to. */
|
||||
uint8_t running_tid; /* Thread assigned ID this client is running on. */
|
||||
uint8_t io_flags; /* Accessed by both main and IO threads, but not modified concurrently */
|
||||
uint8_t read_error; /* Client read error: CLIENT_READ_* macros. */
|
||||
int resp; /* RESP protocol version. Can be 2 or 3. */
|
||||
redisDb *db; /* Pointer to currently SELECTed DB. */
|
||||
robj *name; /* As set by CLIENT SETNAME. */
|
||||
|
|
@ -1226,8 +1267,8 @@ typedef struct client {
|
|||
sds peerid; /* Cached peer ID. */
|
||||
sds sockname; /* Cached connection target address. */
|
||||
listNode *client_list_node; /* list node in client list */
|
||||
listNode *io_thread_client_list_node; /* list node in io thread client list */
|
||||
listNode *postponed_list_node; /* list node within the postponed list */
|
||||
listNode *pending_read_list_node; /* list node in clients pending read list */
|
||||
void *module_blocked_client; /* Pointer to the RedisModuleBlockedClient associated with this
|
||||
* client. This is set in case of module authentication before the
|
||||
* unblocked client is reprocessed to handle reply callbacks. */
|
||||
|
|
@ -1280,6 +1321,20 @@ typedef struct client {
|
|||
#endif
|
||||
} client;
|
||||
|
||||
typedef struct __attribute__((aligned(CACHE_LINE_SIZE))) {
|
||||
uint8_t id; /* The unique ID assigned, if IO_THREADS_MAX_NUM is more
|
||||
* than 256, we should also promote the data type. */
|
||||
pthread_t tid; /* Pthread ID */
|
||||
redisAtomic int paused; /* Paused status for the io thread. */
|
||||
aeEventLoop *el; /* Main event loop of io thread. */
|
||||
list *pending_clients; /* List of clients with pending writes. */
|
||||
list *processing_clients; /* List of clients being processed. */
|
||||
eventNotifier *pending_clients_notifier; /* Used to wake up the loop when write should be performed. */
|
||||
pthread_mutex_t pending_clients_mutex; /* Mutex for pending write list */
|
||||
list *pending_clients_to_main_thread; /* Clients that are waiting to be executed by the main thread. */
|
||||
list *clients; /* IO thread managed clients. */
|
||||
} IOThread;
|
||||
|
||||
/* ACL information */
|
||||
typedef struct aclInfo {
|
||||
long long user_auth_failures; /* Auth failure counts on user level */
|
||||
|
|
@ -1568,6 +1623,7 @@ struct redisServer {
|
|||
int errors_enabled; /* If true, errorstats is enabled, and we will add new errors. */
|
||||
unsigned int lruclock; /* Clock for LRU eviction */
|
||||
volatile sig_atomic_t shutdown_asap; /* Shutdown ordered by signal handler. */
|
||||
volatile sig_atomic_t crashing; /* Server is crashing report. */
|
||||
mstime_t shutdown_mstime; /* Timestamp to limit graceful shutdown. */
|
||||
int last_sig_received; /* Indicates the last SIGNAL received, if any (e.g., SIGINT or SIGTERM). */
|
||||
int shutdown_flags; /* Flags passed to prepareForShutdown(). */
|
||||
|
|
@ -1638,6 +1694,7 @@ struct redisServer {
|
|||
redisAtomic uint64_t next_client_id; /* Next client unique ID. Incremental. */
|
||||
int protected_mode; /* Don't accept external connections. */
|
||||
int io_threads_num; /* Number of IO threads to use. */
|
||||
int io_threads_clients_num[IO_THREADS_MAX_NUM]; /* Number of clients assigned to each IO thread. */
|
||||
int io_threads_do_reads; /* Read and parse from IO threads? */
|
||||
int io_threads_active; /* Is IO threads currently active? */
|
||||
long long events_processed_while_blocked; /* processEventsWhileBlocked() */
|
||||
|
|
@ -1710,8 +1767,8 @@ struct redisServer {
|
|||
long long stat_unexpected_error_replies; /* Number of unexpected (aof-loading, replica to master, etc.) error replies */
|
||||
long long stat_total_error_replies; /* Total number of issued error replies ( command + rejected errors ) */
|
||||
long long stat_dump_payload_sanitizations; /* Number deep dump payloads integrity validations. */
|
||||
long long stat_io_reads_processed; /* Number of read events processed by IO / Main threads */
|
||||
long long stat_io_writes_processed; /* Number of write events processed by IO / Main threads */
|
||||
redisAtomic long long stat_io_reads_processed; /* Number of read events processed by IO / Main threads */
|
||||
redisAtomic long long stat_io_writes_processed; /* Number of write events processed by IO / Main threads */
|
||||
redisAtomic long long stat_total_reads_processed; /* Total number of read events processed */
|
||||
redisAtomic long long stat_total_writes_processed; /* Total number of write events processed */
|
||||
redisAtomic long long stat_client_qbuf_limit_disconnections; /* Total number of clients reached query buf length limit */
|
||||
|
|
@ -2461,11 +2518,6 @@ typedef struct {
|
|||
#define OBJ_HASH_KEY 1
|
||||
#define OBJ_HASH_VALUE 2
|
||||
|
||||
#define IO_THREADS_OP_IDLE 0
|
||||
#define IO_THREADS_OP_READ 1
|
||||
#define IO_THREADS_OP_WRITE 2
|
||||
extern int io_threads_op;
|
||||
|
||||
/* Hash-field data type (of t_hash.c) */
|
||||
typedef mstr hfield;
|
||||
extern mstrKind mstrFieldKind;
|
||||
|
|
@ -2680,9 +2732,6 @@ void whileBlockedCron(void);
|
|||
void blockingOperationStarts(void);
|
||||
void blockingOperationEnds(void);
|
||||
int handleClientsWithPendingWrites(void);
|
||||
int handleClientsWithPendingWritesUsingThreads(void);
|
||||
int handleClientsWithPendingReadsUsingThreads(void);
|
||||
int stopThreadedIOIfNeeded(void);
|
||||
int clientHasPendingReplies(client *c);
|
||||
int updateClientMemUsageAndBucket(client *c);
|
||||
void removeClientFromMemUsageBucket(client *c, int allow_eviction);
|
||||
|
|
@ -2691,13 +2740,32 @@ int writeToClient(client *c, int handler_installed);
|
|||
void linkClient(client *c);
|
||||
void protectClient(client *c);
|
||||
void unprotectClient(client *c);
|
||||
void initThreadedIO(void);
|
||||
client *lookupClientByID(uint64_t id);
|
||||
int authRequired(client *c);
|
||||
void putClientInPendingWriteQueue(client *c);
|
||||
/* reply macros */
|
||||
#define ADD_REPLY_BULK_CBUFFER_STRING_CONSTANT(c, str) addReplyBulkCBuffer(c, str, strlen(str))
|
||||
|
||||
/* iothread.c - the threaded io implementation */
|
||||
void initThreadedIO(void);
|
||||
void killIOThreads(void);
|
||||
void pauseIOThread(int id);
|
||||
void resumeIOThread(int id);
|
||||
void pauseAllIOThreads(void);
|
||||
void resumeAllIOThreads(void);
|
||||
void pauseIOThreadsRange(int start, int end);
|
||||
void resumeIOThreadsRange(int start, int end);
|
||||
int resizeAllIOThreadsEventLoops(size_t newsize);
|
||||
int sendPendingClientsToIOThreads(void);
|
||||
void enqueuePendingClientsToMainThread(client *c, int unbind);
|
||||
void putInPendingClienstForIOThreads(client *c);
|
||||
void handleClientReadError(client *c);
|
||||
void unbindClientFromIOThreadEventLoop(client *c);
|
||||
void processClientsOfAllIOThreads(void);
|
||||
void assignClientToIOThread(client *c);
|
||||
void fetchClientFromIOThread(client *c);
|
||||
int isClientMustHandledByMainThread(client *c);
|
||||
|
||||
/* logreqres.c - logging of requests and responses */
|
||||
void reqresReset(client *c, int free_buf);
|
||||
void reqresSaveClientReplyOffset(client *c);
|
||||
|
|
@ -3901,7 +3969,6 @@ void xorDigest(unsigned char *digest, const void *ptr, size_t len);
|
|||
sds catSubCommandFullname(const char *parent_name, const char *sub_name);
|
||||
void commandAddSubcommand(struct redisCommand *parent, struct redisCommand *subcommand, const char *declared_name);
|
||||
void debugDelay(int usec);
|
||||
void killIOThreads(void);
|
||||
void killThreads(void);
|
||||
void makeThreadKillable(void);
|
||||
void swapMainDbWithTempDb(redisDb *tempDb);
|
||||
|
|
|
|||
37
src/socket.c
37
src/socket.c
|
|
@ -53,11 +53,12 @@ static ConnectionType CT_Socket;
|
|||
* be embedded in different structs, not just client.
|
||||
*/
|
||||
|
||||
static connection *connCreateSocket(void) {
|
||||
static connection *connCreateSocket(struct aeEventLoop *el) {
|
||||
connection *conn = zcalloc(sizeof(connection));
|
||||
conn->type = &CT_Socket;
|
||||
conn->fd = -1;
|
||||
conn->iovcnt = IOV_MAX;
|
||||
conn->el = el;
|
||||
|
||||
return conn;
|
||||
}
|
||||
|
|
@ -72,9 +73,9 @@ static connection *connCreateSocket(void) {
|
|||
* is not in an error state (which is not possible for a socket connection,
|
||||
* but could but possible with other protocols).
|
||||
*/
|
||||
static connection *connCreateAcceptedSocket(int fd, void *priv) {
|
||||
static connection *connCreateAcceptedSocket(struct aeEventLoop *el, int fd, void *priv) {
|
||||
UNUSED(priv);
|
||||
connection *conn = connCreateSocket();
|
||||
connection *conn = connCreateSocket(el);
|
||||
conn->fd = fd;
|
||||
conn->state = CONN_STATE_ACCEPTING;
|
||||
return conn;
|
||||
|
|
@ -93,7 +94,7 @@ static int connSocketConnect(connection *conn, const char *addr, int port, const
|
|||
conn->state = CONN_STATE_CONNECTING;
|
||||
|
||||
conn->conn_handler = connect_handler;
|
||||
aeCreateFileEvent(server.el, conn->fd, AE_WRITABLE,
|
||||
aeCreateFileEvent(conn->el, conn->fd, AE_WRITABLE,
|
||||
conn->type->ae_handler, conn);
|
||||
|
||||
return C_OK;
|
||||
|
|
@ -114,7 +115,7 @@ static void connSocketShutdown(connection *conn) {
|
|||
/* Close the connection and free resources. */
|
||||
static void connSocketClose(connection *conn) {
|
||||
if (conn->fd != -1) {
|
||||
aeDeleteFileEvent(server.el,conn->fd, AE_READABLE | AE_WRITABLE);
|
||||
if (conn->el) aeDeleteFileEvent(conn->el, conn->fd, AE_READABLE | AE_WRITABLE);
|
||||
close(conn->fd);
|
||||
conn->fd = -1;
|
||||
}
|
||||
|
|
@ -190,6 +191,15 @@ static int connSocketAccept(connection *conn, ConnectionCallbackFunc accept_hand
|
|||
return ret;
|
||||
}
|
||||
|
||||
/* Rebind the connection to another event loop, read/write handlers must not
|
||||
* be installed in the current event loop, otherwise it will cause two event
|
||||
* loops to manage the same connection at the same time. */
|
||||
static int connSocketRebindEventLoop(connection *conn, aeEventLoop *el) {
|
||||
serverAssert(!conn->el && !conn->read_handler && !conn->write_handler);
|
||||
conn->el = el;
|
||||
return C_OK;
|
||||
}
|
||||
|
||||
/* Register a write handler, to be called when the connection is writable.
|
||||
* If NULL, the existing handler is removed.
|
||||
*
|
||||
|
|
@ -207,9 +217,9 @@ static int connSocketSetWriteHandler(connection *conn, ConnectionCallbackFunc fu
|
|||
else
|
||||
conn->flags &= ~CONN_FLAG_WRITE_BARRIER;
|
||||
if (!conn->write_handler)
|
||||
aeDeleteFileEvent(server.el,conn->fd,AE_WRITABLE);
|
||||
aeDeleteFileEvent(conn->el,conn->fd,AE_WRITABLE);
|
||||
else
|
||||
if (aeCreateFileEvent(server.el,conn->fd,AE_WRITABLE,
|
||||
if (aeCreateFileEvent(conn->el,conn->fd,AE_WRITABLE,
|
||||
conn->type->ae_handler,conn) == AE_ERR) return C_ERR;
|
||||
return C_OK;
|
||||
}
|
||||
|
|
@ -222,9 +232,9 @@ static int connSocketSetReadHandler(connection *conn, ConnectionCallbackFunc fun
|
|||
|
||||
conn->read_handler = func;
|
||||
if (!conn->read_handler)
|
||||
aeDeleteFileEvent(server.el,conn->fd,AE_READABLE);
|
||||
aeDeleteFileEvent(conn->el,conn->fd,AE_READABLE);
|
||||
else
|
||||
if (aeCreateFileEvent(server.el,conn->fd,
|
||||
if (aeCreateFileEvent(conn->el,conn->fd,
|
||||
AE_READABLE,conn->type->ae_handler,conn) == AE_ERR) return C_ERR;
|
||||
return C_OK;
|
||||
}
|
||||
|
|
@ -250,7 +260,7 @@ static void connSocketEventHandler(struct aeEventLoop *el, int fd, void *clientD
|
|||
conn->state = CONN_STATE_CONNECTED;
|
||||
}
|
||||
|
||||
if (!conn->write_handler) aeDeleteFileEvent(server.el,conn->fd,AE_WRITABLE);
|
||||
if (!conn->write_handler) aeDeleteFileEvent(conn->el, conn->fd, AE_WRITABLE);
|
||||
|
||||
if (!callHandler(conn, conn->conn_handler)) return;
|
||||
conn->conn_handler = NULL;
|
||||
|
|
@ -291,7 +301,6 @@ static void connSocketAcceptHandler(aeEventLoop *el, int fd, void *privdata, int
|
|||
int cport, cfd;
|
||||
int max = server.max_new_conns_per_cycle;
|
||||
char cip[NET_IP_STR_LEN];
|
||||
UNUSED(el);
|
||||
UNUSED(mask);
|
||||
UNUSED(privdata);
|
||||
|
||||
|
|
@ -304,7 +313,7 @@ static void connSocketAcceptHandler(aeEventLoop *el, int fd, void *privdata, int
|
|||
return;
|
||||
}
|
||||
serverLog(LL_VERBOSE,"Accepted %s:%d", cip, cport);
|
||||
acceptCommonHandler(connCreateAcceptedSocket(cfd, NULL),0,cip);
|
||||
acceptCommonHandler(connCreateAcceptedSocket(el,cfd,NULL), 0, cip);
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -397,6 +406,10 @@ static ConnectionType CT_Socket = {
|
|||
.blocking_connect = connSocketBlockingConnect,
|
||||
.accept = connSocketAccept,
|
||||
|
||||
/* event loop */
|
||||
.unbind_event_loop = NULL,
|
||||
.rebind_event_loop = connSocketRebindEventLoop,
|
||||
|
||||
/* IO */
|
||||
.write = connSocketWrite,
|
||||
.writev = connSocketWritev,
|
||||
|
|
|
|||
114
src/tls.c
114
src/tls.c
|
|
@ -75,10 +75,6 @@ static int parseProtocolsConfig(const char *str) {
|
|||
return protocols;
|
||||
}
|
||||
|
||||
/* list of connections with pending data already read from the socket, but not
|
||||
* served to the reader yet. */
|
||||
static list *pending_list = NULL;
|
||||
|
||||
/**
|
||||
* OpenSSL global initialization and locking handling callbacks.
|
||||
* Note that this is only required for OpenSSL < 1.1.0.
|
||||
|
|
@ -144,8 +140,6 @@ static void tlsInit(void) {
|
|||
if (!RAND_poll()) {
|
||||
serverLog(LL_WARNING, "OpenSSL: Failed to seed random number generator.");
|
||||
}
|
||||
|
||||
pending_list = listCreate();
|
||||
}
|
||||
|
||||
static void tlsCleanup(void) {
|
||||
|
|
@ -435,20 +429,21 @@ typedef struct tls_connection {
|
|||
listNode *pending_list_node;
|
||||
} tls_connection;
|
||||
|
||||
static connection *createTLSConnection(int client_side) {
|
||||
static connection *createTLSConnection(struct aeEventLoop *el, int client_side) {
|
||||
SSL_CTX *ctx = redis_tls_ctx;
|
||||
if (client_side && redis_tls_client_ctx)
|
||||
ctx = redis_tls_client_ctx;
|
||||
tls_connection *conn = zcalloc(sizeof(tls_connection));
|
||||
conn->c.type = &CT_TLS;
|
||||
conn->c.fd = -1;
|
||||
conn->c.el = el;
|
||||
conn->c.iovcnt = IOV_MAX;
|
||||
conn->ssl = SSL_new(ctx);
|
||||
return (connection *) conn;
|
||||
}
|
||||
|
||||
static connection *connCreateTLS(void) {
|
||||
return createTLSConnection(1);
|
||||
static connection *connCreateTLS(struct aeEventLoop *el) {
|
||||
return createTLSConnection(el, 1);
|
||||
}
|
||||
|
||||
/* Fetch the latest OpenSSL error and store it in the connection */
|
||||
|
|
@ -468,10 +463,11 @@ static void updateTLSError(tls_connection *conn) {
|
|||
* Callers should use connGetState() and verify the created connection
|
||||
* is not in an error state.
|
||||
*/
|
||||
static connection *connCreateAcceptedTLS(int fd, void *priv) {
|
||||
static connection *connCreateAcceptedTLS(struct aeEventLoop *el, int fd, void *priv) {
|
||||
int require_auth = *(int *)priv;
|
||||
tls_connection *conn = (tls_connection *) createTLSConnection(0);
|
||||
tls_connection *conn = (tls_connection *) createTLSConnection(el, 0);
|
||||
conn->c.fd = fd;
|
||||
conn->c.el = el;
|
||||
conn->c.state = CONN_STATE_ACCEPTING;
|
||||
|
||||
if (!conn->ssl) {
|
||||
|
|
@ -575,17 +571,17 @@ static int updateStateAfterSSLIO(tls_connection *conn, int ret_value, int update
|
|||
}
|
||||
|
||||
static void registerSSLEvent(tls_connection *conn, WantIOType want) {
|
||||
int mask = aeGetFileEvents(server.el, conn->c.fd);
|
||||
int mask = aeGetFileEvents(conn->c.el, conn->c.fd);
|
||||
|
||||
switch (want) {
|
||||
case WANT_READ:
|
||||
if (mask & AE_WRITABLE) aeDeleteFileEvent(server.el, conn->c.fd, AE_WRITABLE);
|
||||
if (!(mask & AE_READABLE)) aeCreateFileEvent(server.el, conn->c.fd, AE_READABLE,
|
||||
if (mask & AE_WRITABLE) aeDeleteFileEvent(conn->c.el, conn->c.fd, AE_WRITABLE);
|
||||
if (!(mask & AE_READABLE)) aeCreateFileEvent(conn->c.el, conn->c.fd, AE_READABLE,
|
||||
tlsEventHandler, conn);
|
||||
break;
|
||||
case WANT_WRITE:
|
||||
if (mask & AE_READABLE) aeDeleteFileEvent(server.el, conn->c.fd, AE_READABLE);
|
||||
if (!(mask & AE_WRITABLE)) aeCreateFileEvent(server.el, conn->c.fd, AE_WRITABLE,
|
||||
if (mask & AE_READABLE) aeDeleteFileEvent(conn->c.el, conn->c.fd, AE_READABLE);
|
||||
if (!(mask & AE_WRITABLE)) aeCreateFileEvent(conn->c.el, conn->c.fd, AE_WRITABLE,
|
||||
tlsEventHandler, conn);
|
||||
break;
|
||||
default:
|
||||
|
|
@ -595,19 +591,42 @@ static void registerSSLEvent(tls_connection *conn, WantIOType want) {
|
|||
}
|
||||
|
||||
static void updateSSLEvent(tls_connection *conn) {
|
||||
int mask = aeGetFileEvents(server.el, conn->c.fd);
|
||||
serverAssert(conn->c.el);
|
||||
int mask = aeGetFileEvents(conn->c.el, conn->c.fd);
|
||||
int need_read = conn->c.read_handler || (conn->flags & TLS_CONN_FLAG_WRITE_WANT_READ);
|
||||
int need_write = conn->c.write_handler || (conn->flags & TLS_CONN_FLAG_READ_WANT_WRITE);
|
||||
|
||||
if (need_read && !(mask & AE_READABLE))
|
||||
aeCreateFileEvent(server.el, conn->c.fd, AE_READABLE, tlsEventHandler, conn);
|
||||
aeCreateFileEvent(conn->c.el, conn->c.fd, AE_READABLE, tlsEventHandler, conn);
|
||||
if (!need_read && (mask & AE_READABLE))
|
||||
aeDeleteFileEvent(server.el, conn->c.fd, AE_READABLE);
|
||||
aeDeleteFileEvent(conn->c.el, conn->c.fd, AE_READABLE);
|
||||
|
||||
if (need_write && !(mask & AE_WRITABLE))
|
||||
aeCreateFileEvent(server.el, conn->c.fd, AE_WRITABLE, tlsEventHandler, conn);
|
||||
aeCreateFileEvent(conn->c.el, conn->c.fd, AE_WRITABLE, tlsEventHandler, conn);
|
||||
if (!need_write && (mask & AE_WRITABLE))
|
||||
aeDeleteFileEvent(server.el, conn->c.fd, AE_WRITABLE);
|
||||
aeDeleteFileEvent(conn->c.el, conn->c.fd, AE_WRITABLE);
|
||||
}
|
||||
|
||||
/* Add a connection to the list of connections with pending data that has
|
||||
* already been read from the socket but has not yet been served to the reader. */
|
||||
static void tlsPendingAdd(tls_connection *conn) {
|
||||
if (!conn->c.el->privdata[1])
|
||||
conn->c.el->privdata[1] = listCreate();
|
||||
|
||||
list *pending_list = conn->c.el->privdata[1];
|
||||
if (!conn->pending_list_node) {
|
||||
listAddNodeTail(pending_list, conn);
|
||||
conn->pending_list_node = listLast(pending_list);
|
||||
}
|
||||
}
|
||||
|
||||
/* Removes a connection from the list of connections with pending data. */
|
||||
static void tlsPendingRemove(tls_connection *conn) {
|
||||
if (conn->pending_list_node) {
|
||||
list *pending_list = conn->c.el->privdata[1];
|
||||
listDelNode(pending_list, conn->pending_list_node);
|
||||
conn->pending_list_node = NULL;
|
||||
}
|
||||
}
|
||||
|
||||
static void tlsHandleEvent(tls_connection *conn, int mask) {
|
||||
|
|
@ -718,13 +737,9 @@ static void tlsHandleEvent(tls_connection *conn, int mask) {
|
|||
* to a list of pending connection that should be handled anyway. */
|
||||
if ((mask & AE_READABLE)) {
|
||||
if (SSL_pending(conn->ssl) > 0) {
|
||||
if (!conn->pending_list_node) {
|
||||
listAddNodeTail(pending_list, conn);
|
||||
conn->pending_list_node = listLast(pending_list);
|
||||
}
|
||||
tlsPendingAdd(conn);
|
||||
} else if (conn->pending_list_node) {
|
||||
listDelNode(pending_list, conn->pending_list_node);
|
||||
conn->pending_list_node = NULL;
|
||||
tlsPendingRemove(conn);
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -734,7 +749,8 @@ static void tlsHandleEvent(tls_connection *conn, int mask) {
|
|||
break;
|
||||
}
|
||||
|
||||
updateSSLEvent(conn);
|
||||
/* The event loop may have been unbound during the event processing above. */
|
||||
if (conn->c.el) updateSSLEvent(conn);
|
||||
}
|
||||
|
||||
static void tlsEventHandler(struct aeEventLoop *el, int fd, void *clientData, int mask) {
|
||||
|
|
@ -748,7 +764,6 @@ static void tlsAcceptHandler(aeEventLoop *el, int fd, void *privdata, int mask)
|
|||
int cport, cfd;
|
||||
int max = server.max_new_tls_conns_per_cycle;
|
||||
char cip[NET_IP_STR_LEN];
|
||||
UNUSED(el);
|
||||
UNUSED(mask);
|
||||
UNUSED(privdata);
|
||||
|
||||
|
|
@ -761,7 +776,7 @@ static void tlsAcceptHandler(aeEventLoop *el, int fd, void *privdata, int mask)
|
|||
return;
|
||||
}
|
||||
serverLog(LL_VERBOSE,"Accepted %s:%d", cip, cport);
|
||||
acceptCommonHandler(connCreateAcceptedTLS(cfd, &server.tls_auth_clients),0,cip);
|
||||
acceptCommonHandler(connCreateAcceptedTLS(el,cfd,&server.tls_auth_clients), 0, cip);
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -806,6 +821,7 @@ static void connTLSClose(connection *conn_) {
|
|||
}
|
||||
|
||||
if (conn->pending_list_node) {
|
||||
list *pending_list = conn->c.el->privdata[1];
|
||||
listDelNode(pending_list, conn->pending_list_node);
|
||||
conn->pending_list_node = NULL;
|
||||
}
|
||||
|
|
@ -863,6 +879,33 @@ static int connTLSConnect(connection *conn_, const char *addr, int port, const c
|
|||
return C_OK;
|
||||
}
|
||||
|
||||
static void connTLSUnbindEventLoop(connection *conn_) {
|
||||
tls_connection *conn = (tls_connection *) conn_;
|
||||
|
||||
/* We need to remove all events from the old event loop. The subsequent
|
||||
* updateSSLEvent() will add the appropriate events to the new event loop. */
|
||||
if (conn->c.el) {
|
||||
int mask = aeGetFileEvents(conn->c.el, conn->c.fd);
|
||||
if (mask & AE_READABLE) aeDeleteFileEvent(conn->c.el, conn->c.fd, AE_READABLE);
|
||||
if (mask & AE_WRITABLE) aeDeleteFileEvent(conn->c.el, conn->c.fd, AE_WRITABLE);
|
||||
|
||||
/* Check if there are pending events and handle accordingly. */
|
||||
int has_pending = conn->pending_list_node != NULL;
|
||||
if (has_pending) tlsPendingRemove(conn);
|
||||
}
|
||||
}
|
||||
|
||||
static int connTLSRebindEventLoop(connection *conn_, aeEventLoop *el) {
|
||||
tls_connection *conn = (tls_connection *) conn_;
|
||||
serverAssert(!conn->c.el && !conn->c.read_handler &&
|
||||
!conn->c.write_handler && !conn->pending_list_node);
|
||||
conn->c.el = el;
|
||||
if (el && SSL_pending(conn->ssl)) tlsPendingAdd(conn);
|
||||
/* Add the appropriate events to the new event loop. */
|
||||
updateSSLEvent((tls_connection *) conn);
|
||||
return C_OK;
|
||||
}
|
||||
|
||||
static int connTLSWrite(connection *conn_, const void *data, size_t data_len) {
|
||||
tls_connection *conn = (tls_connection *) conn_;
|
||||
int ret;
|
||||
|
|
@ -1044,16 +1087,19 @@ static const char *connTLSGetType(connection *conn_) {
|
|||
return CONN_TYPE_TLS;
|
||||
}
|
||||
|
||||
static int tlsHasPendingData(void) {
|
||||
static int tlsHasPendingData(struct aeEventLoop *el) {
|
||||
list *pending_list = el->privdata[1];
|
||||
if (!pending_list)
|
||||
return 0;
|
||||
return listLength(pending_list) > 0;
|
||||
}
|
||||
|
||||
static int tlsProcessPendingData(void) {
|
||||
static int tlsProcessPendingData(struct aeEventLoop *el) {
|
||||
listIter li;
|
||||
listNode *ln;
|
||||
|
||||
list *pending_list = el->privdata[1];
|
||||
if (!pending_list) return 0;
|
||||
int processed = listLength(pending_list);
|
||||
listRewind(pending_list,&li);
|
||||
while((ln = listNext(&li))) {
|
||||
|
|
@ -1114,6 +1160,10 @@ static ConnectionType CT_TLS = {
|
|||
.blocking_connect = connTLSBlockingConnect,
|
||||
.accept = connTLSAccept,
|
||||
|
||||
/* event loop */
|
||||
.unbind_event_loop = connTLSUnbindEventLoop,
|
||||
.rebind_event_loop = connTLSRebindEventLoop,
|
||||
|
||||
/* IO */
|
||||
.read = connTLSRead,
|
||||
.write = connTLSWrite,
|
||||
|
|
|
|||
|
|
@ -253,6 +253,7 @@ void trackingRememberKeys(client *tracking, client *executing) {
|
|||
* - Following a flush command, to send a single RESP NULL to indicate
|
||||
* that all keys are now invalid. */
|
||||
void sendTrackingMessage(client *c, char *keyname, size_t keylen, int proto) {
|
||||
int paused = 0;
|
||||
uint64_t old_flags = c->flags;
|
||||
c->flags |= CLIENT_PUSHING;
|
||||
|
||||
|
|
@ -275,6 +276,11 @@ void sendTrackingMessage(client *c, char *keyname, size_t keylen, int proto) {
|
|||
if (!(old_flags & CLIENT_PUSHING)) c->flags &= ~CLIENT_PUSHING;
|
||||
c = redir;
|
||||
using_redirection = 1;
|
||||
/* Start to touch another client data. */
|
||||
if (c->running_tid != IOTHREAD_MAIN_THREAD_ID) {
|
||||
pauseIOThread(c->running_tid);
|
||||
paused = 1;
|
||||
}
|
||||
old_flags = c->flags;
|
||||
c->flags |= CLIENT_PUSHING;
|
||||
}
|
||||
|
|
@ -296,7 +302,7 @@ void sendTrackingMessage(client *c, char *keyname, size_t keylen, int proto) {
|
|||
* it since RESP2 does not support push messages in the same
|
||||
* connection. */
|
||||
if (!(old_flags & CLIENT_PUSHING)) c->flags &= ~CLIENT_PUSHING;
|
||||
return;
|
||||
goto done;
|
||||
}
|
||||
|
||||
/* Send the "value" part, which is the array of keys. */
|
||||
|
|
@ -308,6 +314,17 @@ void sendTrackingMessage(client *c, char *keyname, size_t keylen, int proto) {
|
|||
}
|
||||
updateClientMemUsageAndBucket(c);
|
||||
if (!(old_flags & CLIENT_PUSHING)) c->flags &= ~CLIENT_PUSHING;
|
||||
|
||||
done:
|
||||
if (paused) {
|
||||
if (clientHasPendingReplies(c)) {
|
||||
serverAssert(!(c->flags & CLIENT_PENDING_WRITE));
|
||||
/* Actually we install write handler of client which is in IO thread
|
||||
* event loop, it is safe since the io thread is paused */
|
||||
connSetWriteHandler(c->conn, sendReplyToClient);
|
||||
}
|
||||
resumeIOThread(c->running_tid);
|
||||
}
|
||||
}
|
||||
|
||||
/* This function is called when a key is modified in Redis and in the case
|
||||
|
|
|
|||
17
src/unix.c
17
src/unix.c
|
|
@ -74,18 +74,19 @@ static int connUnixListen(connListener *listener) {
|
|||
return C_OK;
|
||||
}
|
||||
|
||||
static connection *connCreateUnix(void) {
|
||||
static connection *connCreateUnix(struct aeEventLoop *el) {
|
||||
connection *conn = zcalloc(sizeof(connection));
|
||||
conn->type = &CT_Unix;
|
||||
conn->fd = -1;
|
||||
conn->iovcnt = IOV_MAX;
|
||||
conn->el = el;
|
||||
|
||||
return conn;
|
||||
}
|
||||
|
||||
static connection *connCreateAcceptedUnix(int fd, void *priv) {
|
||||
static connection *connCreateAcceptedUnix(struct aeEventLoop *el, int fd, void *priv) {
|
||||
UNUSED(priv);
|
||||
connection *conn = connCreateUnix();
|
||||
connection *conn = connCreateUnix(el);
|
||||
conn->fd = fd;
|
||||
conn->state = CONN_STATE_ACCEPTING;
|
||||
return conn;
|
||||
|
|
@ -107,7 +108,7 @@ static void connUnixAcceptHandler(aeEventLoop *el, int fd, void *privdata, int m
|
|||
return;
|
||||
}
|
||||
serverLog(LL_VERBOSE,"Accepted connection to %s", server.unixsocket);
|
||||
acceptCommonHandler(connCreateAcceptedUnix(cfd, NULL),CLIENT_UNIX_SOCKET,NULL);
|
||||
acceptCommonHandler(connCreateAcceptedUnix(el, cfd, NULL),CLIENT_UNIX_SOCKET,NULL);
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -123,6 +124,10 @@ static int connUnixAccept(connection *conn, ConnectionCallbackFunc accept_handle
|
|||
return connectionTypeTcp()->accept(conn, accept_handler);
|
||||
}
|
||||
|
||||
static int connUnixRebindEventLoop(connection *conn, aeEventLoop *el) {
|
||||
return connectionTypeTcp()->rebind_event_loop(conn, el);
|
||||
}
|
||||
|
||||
static int connUnixWrite(connection *conn, const void *data, size_t data_len) {
|
||||
return connectionTypeTcp()->write(conn, data, data_len);
|
||||
}
|
||||
|
|
@ -186,6 +191,10 @@ static ConnectionType CT_Unix = {
|
|||
.blocking_connect = NULL,
|
||||
.accept = connUnixAccept,
|
||||
|
||||
/* event loop */
|
||||
.unbind_event_loop = NULL,
|
||||
.rebind_event_loop = connUnixRebindEventLoop,
|
||||
|
||||
/* IO */
|
||||
.write = connUnixWrite,
|
||||
.writev = connUnixWritev,
|
||||
|
|
|
|||
|
|
@ -156,6 +156,11 @@ test "Shutting down master waits for replica then fails" {
|
|||
set rd2 [redis_deferring_client -1]
|
||||
$rd1 shutdown
|
||||
$rd2 shutdown
|
||||
wait_for_condition 100 10 {
|
||||
[llength [regexp -all -inline {cmd=shutdown} [$master client list]]] eq 2
|
||||
} else {
|
||||
fail "shutdown did not arrive"
|
||||
}
|
||||
set info_clients [$master info clients]
|
||||
assert_match "*connected_clients:3*" $info_clients
|
||||
assert_match "*blocked_clients:2*" $info_clients
|
||||
|
|
@ -209,6 +214,11 @@ test "Shutting down master waits for replica then aborted" {
|
|||
set rd2 [redis_deferring_client -1]
|
||||
$rd1 shutdown
|
||||
$rd2 shutdown
|
||||
wait_for_condition 100 10 {
|
||||
[llength [regexp -all -inline {cmd=shutdown} [$master client list]]] eq 2
|
||||
} else {
|
||||
fail "shutdown did not arrive"
|
||||
}
|
||||
set info_clients [$master info clients]
|
||||
assert_match "*connected_clients:3*" $info_clients
|
||||
assert_match "*blocked_clients:2*" $info_clients
|
||||
|
|
|
|||
|
|
@ -698,6 +698,16 @@ proc latencyrstat_percentiles {cmd r} {
|
|||
}
|
||||
}
|
||||
|
||||
proc get_io_thread_clients {id {client r}} {
|
||||
set pattern "io_thread_$id:clients=(\[0-9\]+)"
|
||||
set info [$client info threads]
|
||||
if {[regexp $pattern $info _ value]} {
|
||||
return $value
|
||||
} else {
|
||||
return -1
|
||||
}
|
||||
}
|
||||
|
||||
proc generate_fuzzy_traffic_on_key {key type duration} {
|
||||
# Commands per type, blocking commands removed
|
||||
# TODO: extract these from COMMAND DOCS, and improve to include other types
|
||||
|
|
|
|||
|
|
@ -108,7 +108,11 @@ start_server {} {
|
|||
$rr write [join [list "*1\r\n\$$maxmemory_clients_actual\r\n" [string repeat v $maxmemory_clients_actual]] ""]
|
||||
$rr flush
|
||||
} e
|
||||
assert {![client_exists $cname]}
|
||||
wait_for_condition 100 10 {
|
||||
![client_exists $cname]
|
||||
} else {
|
||||
fail "Failed to evict client"
|
||||
}
|
||||
$rr close
|
||||
|
||||
# Restore settings
|
||||
|
|
@ -360,6 +364,13 @@ start_server {} {
|
|||
resume_process $server_pid
|
||||
r ping ;# make sure a full event loop cycle is processed before issuing CLIENT LIST
|
||||
|
||||
# wait for get commands to be processed
|
||||
wait_for_condition 100 10 {
|
||||
[expr {[regexp {calls=(\d+)} [cmdrstat get r] -> calls] ? $calls : 0}] >= 2
|
||||
} else {
|
||||
fail "get did not arrive"
|
||||
}
|
||||
|
||||
# Validate obuf-clients were disconnected (because of obuf limit)
|
||||
catch {client_field obuf-client1 name} e
|
||||
assert_match {no client named obuf-client1 found*} $e
|
||||
|
|
@ -367,7 +378,9 @@ start_server {} {
|
|||
assert_match {no client named obuf-client2 found*} $e
|
||||
|
||||
# Validate qbuf-client is still connected and wasn't evicted
|
||||
assert_equal [client_field qbuf-client name] {qbuf-client}
|
||||
if {[lindex [r config get io-threads] 1] == 1} {
|
||||
assert_equal [client_field qbuf-client name] {qbuf-client}
|
||||
}
|
||||
|
||||
$rr1 close
|
||||
$rr2 close
|
||||
|
|
@ -404,8 +417,11 @@ start_server {} {
|
|||
|
||||
# Decrease maxmemory_clients and expect client eviction
|
||||
r config set maxmemory-clients [expr $maxmemory_clients / 2]
|
||||
set connected_clients [llength [lsearch -all [split [string trim [r client list]] "\r\n"] *name=client*]]
|
||||
assert {$connected_clients > 0 && $connected_clients < $client_count}
|
||||
wait_for_condition 200 10 {
|
||||
[llength [regexp -all -inline {name=client} [r client list]]] < $client_count
|
||||
} else {
|
||||
fail "Failed to evict clients"
|
||||
}
|
||||
|
||||
foreach rr $rrs {$rr close}
|
||||
}
|
||||
|
|
@ -463,8 +479,11 @@ start_server {} {
|
|||
assert {$total_client_mem <= $maxmemory_clients}
|
||||
|
||||
# Make sure we have only half of our clients now
|
||||
set connected_clients [llength [lsearch -all [split [string trim [r client list]] "\r\n"] *name=client*]]
|
||||
assert {$connected_clients == [expr $client_count / 2]}
|
||||
wait_for_condition 200 100 {
|
||||
[llength [regexp -all -inline {name=client} [r client list]]] == $client_count / 2
|
||||
} else {
|
||||
fail "Failed to evict clients"
|
||||
}
|
||||
|
||||
# Restore the reply buffer resize to default
|
||||
r debug replybuffer resizing 1
|
||||
|
|
@ -519,7 +538,8 @@ start_server {} {
|
|||
foreach size [lreverse $sizes] {
|
||||
set control_mem [client_field control tot-mem]
|
||||
set total_mem [expr $total_mem - $clients_per_size * $size]
|
||||
r config set maxmemory-clients [expr $total_mem + $control_mem]
|
||||
# allow some tolerance when using io threads
|
||||
r config set maxmemory-clients [expr $total_mem + $control_mem + 1000]
|
||||
set clients [split [string trim [r client list]] "\r\n"]
|
||||
# Verify only relevant clients were evicted
|
||||
for {set i 0} {$i < [llength $sizes]} {incr i} {
|
||||
|
|
|
|||
|
|
@ -313,7 +313,7 @@ start_server {tags {"info" "external:skip"}} {
|
|||
assert_lessthan $cycle2 [expr $cycle1+10] ;# we expect 2 or 3 cycles here, but allow some tolerance
|
||||
if {$::verbose} { puts "eventloop metrics el_sum1: $el_sum1, el_sum2: $el_sum2" }
|
||||
assert_morethan $el_sum2 $el_sum1
|
||||
assert_lessthan $el_sum2 [expr $el_sum1+30000] ;# we expect roughly 100ms here, but allow some tolerance
|
||||
assert_lessthan $el_sum2 [expr $el_sum1+100000] ;# we expect roughly 100ms here, but allow some tolerance
|
||||
if {$::verbose} { puts "eventloop metrics cmd_sum1: $cmd_sum1, cmd_sum2: $cmd_sum2" }
|
||||
assert_morethan $cmd_sum2 $cmd_sum1
|
||||
assert_lessthan $cmd_sum2 [expr $cmd_sum1+15000] ;# we expect about tens of ms here, but allow some tolerance
|
||||
|
|
|
|||
|
|
@ -6,8 +6,13 @@ start_server {tags {"introspection"}} {
|
|||
}
|
||||
|
||||
test {CLIENT LIST} {
|
||||
r client list
|
||||
} {id=* addr=*:* laddr=*:* fd=* name=* age=* idle=* flags=N db=* sub=0 psub=0 ssub=0 multi=-1 watch=0 qbuf=26 qbuf-free=* argv-mem=* multi-mem=0 rbs=* rbp=* obl=0 oll=0 omem=0 tot-mem=* events=r cmd=client|list user=* redir=-1 resp=*}
|
||||
set client_list [r client list]
|
||||
if {[lindex [r config get io-threads] 1] == 1} {
|
||||
assert_match {id=* addr=*:* laddr=*:* fd=* name=* age=* idle=* flags=N db=* sub=0 psub=0 ssub=0 multi=-1 watch=0 qbuf=26 qbuf-free=* argv-mem=* multi-mem=0 rbs=* rbp=* obl=0 oll=0 omem=0 tot-mem=* events=r cmd=client|list user=* redir=-1 resp=*} $client_list
|
||||
} else {
|
||||
assert_match {id=* addr=*:* laddr=*:* fd=* name=* age=* idle=* flags=N db=* sub=0 psub=0 ssub=0 multi=-1 watch=0 qbuf=0 qbuf-free=* argv-mem=* multi-mem=0 rbs=* rbp=* obl=0 oll=0 omem=0 tot-mem=* events=r cmd=client|list user=* redir=-1 resp=*} $client_list
|
||||
}
|
||||
}
|
||||
|
||||
test {CLIENT LIST with IDs} {
|
||||
set myid [r client id]
|
||||
|
|
@ -16,8 +21,13 @@ start_server {tags {"introspection"}} {
|
|||
}
|
||||
|
||||
test {CLIENT INFO} {
|
||||
r client info
|
||||
} {id=* addr=*:* laddr=*:* fd=* name=* age=* idle=* flags=N db=* sub=0 psub=0 ssub=0 multi=-1 watch=0 qbuf=26 qbuf-free=* argv-mem=* multi-mem=0 rbs=* rbp=* obl=0 oll=0 omem=0 tot-mem=* events=r cmd=client|info user=* redir=-1 resp=*}
|
||||
set client [r client info]
|
||||
if {[lindex [r config get io-threads] 1] == 1} {
|
||||
assert_match {id=* addr=*:* laddr=*:* fd=* name=* age=* idle=* flags=N db=* sub=0 psub=0 ssub=0 multi=-1 watch=0 qbuf=26 qbuf-free=* argv-mem=* multi-mem=0 rbs=* rbp=* obl=0 oll=0 omem=0 tot-mem=* events=r cmd=client|info user=* redir=-1 resp=*} $client
|
||||
} else {
|
||||
assert_match {id=* addr=*:* laddr=*:* fd=* name=* age=* idle=* flags=N db=* sub=0 psub=0 ssub=0 multi=-1 watch=0 qbuf=0 qbuf-free=* argv-mem=* multi-mem=0 rbs=* rbp=* obl=0 oll=0 omem=0 tot-mem=* events=r cmd=client|info user=* redir=-1 resp=*} $client
|
||||
}
|
||||
}
|
||||
|
||||
test {CLIENT KILL with illegal arguments} {
|
||||
assert_error "ERR wrong number of arguments for 'client|kill' command" {r client kill}
|
||||
|
|
@ -86,6 +96,11 @@ start_server {tags {"introspection"}} {
|
|||
assert {$connected_clients >= 3}
|
||||
set res [r client kill skipme yes]
|
||||
assert {$res == $connected_clients - 1}
|
||||
wait_for_condition 1000 10 {
|
||||
[s connected_clients] eq 1
|
||||
} else {
|
||||
fail "Can't kill all clients except the current one"
|
||||
}
|
||||
|
||||
# Kill all clients, including `me`
|
||||
set rd3 [redis_deferring_client]
|
||||
|
|
@ -304,6 +319,9 @@ start_server {tags {"introspection"}} {
|
|||
$rd read ; # Discard the OK
|
||||
|
||||
$bc blpop mylist 0
|
||||
# make sure the blpop arrives first
|
||||
$bc flush
|
||||
after 100
|
||||
wait_for_blocked_clients_count 1
|
||||
r lpush mylist 1
|
||||
wait_for_blocked_clients_count 0
|
||||
|
|
@ -904,3 +922,62 @@ test {CONFIG REWRITE handles alias config properly} {
|
|||
assert_equal [r config get hash-max-listpack-entries] {hash-max-listpack-entries 100}
|
||||
}
|
||||
} {} {external:skip}
|
||||
|
||||
test {IO threads client number} {
|
||||
start_server {overrides {io-threads 2} tags {external:skip}} {
|
||||
set iothread_clients [get_io_thread_clients 1]
|
||||
assert_equal $iothread_clients [s connected_clients]
|
||||
assert_equal [get_io_thread_clients 0] 0
|
||||
|
||||
r script debug yes ; # Transfer to main thread
|
||||
assert_equal [get_io_thread_clients 0] 1
|
||||
assert_equal [get_io_thread_clients 1] [expr $iothread_clients - 1]
|
||||
|
||||
set iothread_clients [get_io_thread_clients 1]
|
||||
set rd1 [redis_deferring_client]
|
||||
set rd2 [redis_deferring_client]
|
||||
assert_equal [get_io_thread_clients 1] [expr $iothread_clients + 2]
|
||||
$rd1 close
|
||||
$rd2 close
|
||||
wait_for_condition 1000 10 {
|
||||
[get_io_thread_clients 1] eq $iothread_clients
|
||||
} else {
|
||||
fail "Fail to close clients of io thread 1"
|
||||
}
|
||||
assert_equal [get_io_thread_clients 0] 1
|
||||
|
||||
r script debug no ; # Transfer to io thread
|
||||
assert_equal [get_io_thread_clients 0] 0
|
||||
assert_equal [get_io_thread_clients 1] [expr $iothread_clients + 1]
|
||||
}
|
||||
}
|
||||
|
||||
test {Clients are evenly distributed among io threads} {
|
||||
start_server {overrides {io-threads 4} tags {external:skip}} {
|
||||
set cur_clients [s connected_clients]
|
||||
assert_equal $cur_clients 1
|
||||
global rdclients
|
||||
for {set i 1} {$i < 9} {incr i} {
|
||||
set rdclients($i) [redis_deferring_client]
|
||||
}
|
||||
for {set i 1} {$i <= 3} {incr i} {
|
||||
assert_equal [get_io_thread_clients $i] 3
|
||||
}
|
||||
|
||||
$rdclients(3) close
|
||||
$rdclients(4) close
|
||||
wait_for_condition 1000 10 {
|
||||
[get_io_thread_clients 1] eq 2 &&
|
||||
[get_io_thread_clients 2] eq 2 &&
|
||||
[get_io_thread_clients 3] eq 3
|
||||
} else {
|
||||
fail "Fail to close clients"
|
||||
}
|
||||
|
||||
set $rdclients(3) [redis_deferring_client]
|
||||
set $rdclients(4) [redis_deferring_client]
|
||||
for {set i 1} {$i <= 3} {incr i} {
|
||||
assert_equal [get_io_thread_clients $i] 3
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -10,6 +10,7 @@ start_server {tags {"lazyfree"}} {
|
|||
set peak_mem [s used_memory]
|
||||
assert {[r unlink myset] == 1}
|
||||
assert {$peak_mem > $orig_mem+1000000}
|
||||
reconnect ;# free the memory of reused argv of client
|
||||
wait_for_condition 50 100 {
|
||||
[s used_memory] < $peak_mem &&
|
||||
[s used_memory] < $orig_mem*2
|
||||
|
|
@ -32,6 +33,7 @@ start_server {tags {"lazyfree"}} {
|
|||
set peak_mem [s used_memory]
|
||||
r flushdb async
|
||||
assert {$peak_mem > $orig_mem+1000000}
|
||||
reconnect ;# free the memory of reused argv of client
|
||||
wait_for_condition 50 100 {
|
||||
[s used_memory] < $peak_mem &&
|
||||
[s used_memory] < $orig_mem*2
|
||||
|
|
|
|||
|
|
@ -29,7 +29,11 @@ start_server {tags {"maxmemory" "external:skip"}} {
|
|||
set dbsize [r dbsize]
|
||||
|
||||
if $client_eviction {
|
||||
return [expr $evicted_clients > 0 && $evicted_keys == 0 && $dbsize == 50]
|
||||
if {[lindex [r config get io-threads] 1] == 1} {
|
||||
return [expr $evicted_clients > 0 && $evicted_keys == 0 && $dbsize == 50]
|
||||
} else {
|
||||
return [expr $evicted_clients >= 0 && $evicted_keys >= 0 && $dbsize <= 50]
|
||||
}
|
||||
} else {
|
||||
return [expr $evicted_clients == 0 && $evicted_keys > 0 && $dbsize < 50]
|
||||
}
|
||||
|
|
|
|||
|
|
@ -420,7 +420,10 @@ run_solo {defrag} {
|
|||
$rd_pubsub read ; # Discard subscribe replies
|
||||
$rd_pubsub ssubscribe $channel_name
|
||||
$rd_pubsub read ; # Discard ssubscribe replies
|
||||
$rd set k$j $channel_name
|
||||
# Pub/Sub clients are handled in the main thread, so their memory is
|
||||
# allocated there. Using the SETBIT command avoids the main thread
|
||||
# referencing argv from IO threads.
|
||||
$rd setbit k$j [expr {[string length $channel_name] * 8}] 1
|
||||
$rd read ; # Discard set replies
|
||||
}
|
||||
|
||||
|
|
@ -583,6 +586,123 @@ run_solo {defrag} {
|
|||
}
|
||||
}
|
||||
|
||||
test "Active defrag for argv retained by the main thread from IO thread: $type" {
|
||||
r flushdb
|
||||
r config set hz 100
|
||||
r config set activedefrag no
|
||||
wait_for_defrag_stop 500 100
|
||||
r config resetstat
|
||||
set io_threads [lindex [r config get io-threads] 1]
|
||||
if {$io_threads == 1} {
|
||||
r config set active-defrag-threshold-lower 5
|
||||
} else {
|
||||
r config set active-defrag-threshold-lower 10
|
||||
}
|
||||
r config set active-defrag-cycle-min 65
|
||||
r config set active-defrag-cycle-max 75
|
||||
r config set active-defrag-ignore-bytes 1000kb
|
||||
r config set maxmemory 0
|
||||
|
||||
# Create some clients so that they are distributed among different io threads.
|
||||
set clients {}
|
||||
for {set i 0} {$i < 8} {incr i} {
|
||||
lappend clients [redis_client]
|
||||
}
|
||||
|
||||
# Populate memory with interleaving key pattern of same size
|
||||
set dummy "[string repeat x 400]"
|
||||
set n 10000
|
||||
for {set i 0} {$i < [llength $clients]} {incr i} {
|
||||
set rr [lindex $clients $i]
|
||||
for {set j 0} {$j < $n} {incr j} {
|
||||
$rr set "k$i-$j" $dummy
|
||||
}
|
||||
}
|
||||
|
||||
# If io-threads is enable, verify that memory allocation is not from the main thread.
|
||||
if {$io_threads != 1} {
|
||||
# At least make sure that bin 448 is created in the main thread's arena.
|
||||
r set k dummy
|
||||
r del k
|
||||
|
||||
# We created 10000 string keys of 400 bytes each for each client, so when the memory
|
||||
# allocation for the 448 bin in the main thread is significantly smaller than this,
|
||||
# we can conclude that the memory allocation is not coming from it.
|
||||
set malloc_stats [r memory malloc-stats]
|
||||
if {[regexp {(?s)arenas\[0\]:.*?448[ ]+[\d]+[ ]+([\d]+)[ ]} $malloc_stats - allocated]} {
|
||||
# Ensure the allocation for bin 448 in the main thread’s arena
|
||||
# is far less than 4375k (10000 * 448 bytes).
|
||||
assert_lessthan $allocated 200000
|
||||
} else {
|
||||
fail "Failed to get the main thread's malloc stats."
|
||||
}
|
||||
}
|
||||
|
||||
after 120 ;# serverCron only updates the info once in 100ms
|
||||
if {$::verbose} {
|
||||
puts "used [s allocator_allocated]"
|
||||
puts "rss [s allocator_active]"
|
||||
puts "frag [s allocator_frag_ratio]"
|
||||
puts "frag_bytes [s allocator_frag_bytes]"
|
||||
}
|
||||
assert_lessthan [s allocator_frag_ratio] 1.05
|
||||
|
||||
# Delete keys with even indices to create fragmentation.
|
||||
for {set i 0} {$i < [llength $clients]} {incr i} {
|
||||
set rd [lindex $clients $i]
|
||||
for {set j 0} {$j < $n} {incr j 2} {
|
||||
$rd del "k$i-$j"
|
||||
}
|
||||
}
|
||||
for {set i 0} {$i < [llength $clients]} {incr i} {
|
||||
[lindex $clients $i] close
|
||||
}
|
||||
|
||||
after 120 ;# serverCron only updates the info once in 100ms
|
||||
if {$::verbose} {
|
||||
puts "used [s allocator_allocated]"
|
||||
puts "rss [s allocator_active]"
|
||||
puts "frag [s allocator_frag_ratio]"
|
||||
puts "frag_bytes [s allocator_frag_bytes]"
|
||||
}
|
||||
assert_morethan [s allocator_frag_ratio] 1.4
|
||||
|
||||
catch {r config set activedefrag yes} e
|
||||
if {[r config get activedefrag] eq "activedefrag yes"} {
|
||||
|
||||
# wait for the active defrag to start working (decision once a second)
|
||||
wait_for_condition 50 100 {
|
||||
[s total_active_defrag_time] ne 0
|
||||
} else {
|
||||
after 120 ;# serverCron only updates the info once in 100ms
|
||||
puts [r info memory]
|
||||
puts [r info stats]
|
||||
puts [r memory malloc-stats]
|
||||
fail "defrag not started."
|
||||
}
|
||||
|
||||
# wait for the active defrag to stop working
|
||||
wait_for_defrag_stop 500 100
|
||||
|
||||
# test the fragmentation is lower
|
||||
after 120 ;# serverCron only updates the info once in 100ms
|
||||
if {$::verbose} {
|
||||
puts "used [s allocator_allocated]"
|
||||
puts "rss [s allocator_active]"
|
||||
puts "frag [s allocator_frag_ratio]"
|
||||
puts "frag_bytes [s allocator_frag_bytes]"
|
||||
}
|
||||
|
||||
if {$io_threads == 1} {
|
||||
assert_lessthan_equal [s allocator_frag_ratio] 1.05
|
||||
} else {
|
||||
# TODO: When multithreading is enabled, argv may be created in the io thread
|
||||
# and kept in the main thread, which can cause fragmentation to become worse.
|
||||
assert_lessthan_equal [s allocator_frag_ratio] 1.1
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if {$type eq "standalone"} { ;# skip in cluster mode
|
||||
test "Active defrag big list: $type" {
|
||||
r flushdb
|
||||
|
|
|
|||
|
|
@ -130,7 +130,12 @@ foreach call_type {nested normal} {
|
|||
$rd flush
|
||||
|
||||
# make sure we get BUSY error, and that we didn't get it too early
|
||||
assert_error {*BUSY Slow module operation*} {r ping}
|
||||
wait_for_condition 50 100 {
|
||||
([catch {r ping} reply] == 1) &&
|
||||
([string match {*BUSY Slow module operation*} $reply])
|
||||
} else {
|
||||
fail "Failed waiting for busy slow response"
|
||||
}
|
||||
assert_morethan_equal [expr [clock clicks -milliseconds]-$start] $busy_time_limit
|
||||
|
||||
# abort the blocking operation
|
||||
|
|
|
|||
|
|
@ -85,6 +85,11 @@ start_server {tags {"pubsub network"}} {
|
|||
set rd1 [redis_deferring_client]
|
||||
assert_equal {1 2 3} [subscribe $rd1 {chan1 chan2 chan3}]
|
||||
unsubscribe $rd1
|
||||
wait_for_condition 100 10 {
|
||||
[regexp {cmd=unsubscribe} [r client list]] eq 1
|
||||
} else {
|
||||
fail "unsubscribe did not arrive"
|
||||
}
|
||||
assert_equal 0 [r publish chan1 hello]
|
||||
assert_equal 0 [r publish chan2 hello]
|
||||
assert_equal 0 [r publish chan3 hello]
|
||||
|
|
@ -158,6 +163,11 @@ start_server {tags {"pubsub network"}} {
|
|||
set rd1 [redis_deferring_client]
|
||||
assert_equal {1 2 3} [psubscribe $rd1 {chan1.* chan2.* chan3.*}]
|
||||
punsubscribe $rd1
|
||||
wait_for_condition 100 10 {
|
||||
[regexp {cmd=punsubscribe} [r client list]] eq 1
|
||||
} else {
|
||||
fail "punsubscribe did not arrive"
|
||||
}
|
||||
assert_equal 0 [r publish chan1.hi hello]
|
||||
assert_equal 0 [r publish chan2.hi hello]
|
||||
assert_equal 0 [r publish chan3.hi hello]
|
||||
|
|
|
|||
|
|
@ -46,6 +46,11 @@ start_server {tags {"pubsubshard external:skip"}} {
|
|||
assert_equal {2} [ssubscribe $rd1 {chan2}]
|
||||
assert_equal {3} [ssubscribe $rd1 {chan3}]
|
||||
sunsubscribe $rd1
|
||||
wait_for_condition 100 10 {
|
||||
[regexp {cmd=sunsubscribe} [r client list]] eq 1
|
||||
} else {
|
||||
fail "sunsubscribe did not arrive"
|
||||
}
|
||||
assert_equal 0 [r SPUBLISH chan1 hello]
|
||||
assert_equal 0 [r SPUBLISH chan2 hello]
|
||||
assert_equal 0 [r SPUBLISH chan3 hello]
|
||||
|
|
|
|||
|
|
@ -166,7 +166,12 @@ start_server {tags {"querybuf"}} {
|
|||
# The client executing the command is currently using the reusable query buffer,
|
||||
# so the size shown is that of the reusable query buffer. It will be returned
|
||||
# to the reusable query buffer after command execution.
|
||||
assert_match {*qbuf=26 qbuf-free=* cmd=client|list *} $res
|
||||
# Note that if IO threads are enabled, the reusable query buffer will be dereferenced earlier.
|
||||
if {[lindex [r config get io-threads] 1] == 1} {
|
||||
assert_match {*qbuf=26 qbuf-free=* cmd=client|list *} $res
|
||||
} else {
|
||||
assert_match {*qbuf=0 qbuf-free=* cmd=client|list *} $res
|
||||
}
|
||||
|
||||
$rd close
|
||||
}
|
||||
|
|
|
|||
|
|
@ -1100,6 +1100,11 @@ foreach {pop} {BLPOP BLMPOP_LEFT} {
|
|||
$watching_client get somekey{t}
|
||||
$watching_client read
|
||||
$watching_client exec
|
||||
wait_for_condition 100 10 {
|
||||
[regexp {cmd=exec} [r client list]] eq 1
|
||||
} else {
|
||||
fail "exec did not arrive"
|
||||
}
|
||||
# Blocked BLPOPLPUSH may create problems, unblock it.
|
||||
r lpush srclist{t} element
|
||||
set res [$watching_client read]
|
||||
|
|
|
|||
Loading…
Reference in New Issue