gvsafronov/futriix
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
8,887 Commits 1 Branch 0 Tags
Commit Graph

438 Commits

Author SHA1 Message Date
antirez
4e10311fe5 Threaded IO: process read queue before stopping threads. 2019-05-06 18:02:51 +02:00
antirez
b9b29641cf Threaded IO: read side WIP 3. 2019-05-06 18:02:51 +02:00
antirez
1c3b585e78 Threaded IO: read side WIP 2. 2019-05-06 18:02:51 +02:00
antirez
7bae8afef8 Threaded IO: read side WIP. 2019-05-06 18:02:51 +02:00
antirez
793134d0e6 Threaded IO: hide more debugging printfs under conditional. 2019-05-06 18:02:51 +02:00
antirez
dc4d13e751 Threaded IO: make num of I/O threads configurable. 2019-05-06 18:02:51 +02:00
antirez
79626ca340 Threaded IO: use main thread if num of threads is 1. 2019-05-06 18:02:51 +02:00
Ubuntu
0353d13321 Threaded IO: stop threads when no longer needed + C11 in Makefile. 
Now threads are stopped even when the connections drop immediately to
zero, not allowing the networking code to detect the condition and stop
the threads. serverCron() will handle that.
 2019-05-06 18:02:51 +02:00
antirez
444b5f2af5 Threaded IO: 3rd version: use the mutex only to stop the thread. 2019-05-06 18:02:51 +02:00
antirez
df0a28f661 Threaded IO: second attempt without signaling conditions. 2019-05-06 18:02:51 +02:00
antirez
c1cb2ae695 Threaded IO: allow to disable debug printf. 2019-05-06 18:02:51 +02:00
antirez
62261aa905 Threaded IO: implement handleClientsWithPendingWritesUsingThreads(). 
This is just an experiment for now, there are a couple of race
conditions, mostly harmless for the performance gain experiment that
this commit represents so far.

The general idea here is to take Redis single threaded and instead
fan-out on expansive kernel calls: write(2) in this case, but the same
concept could be easily implemented for read(2) and protcol parsing.

However just threading writes like in this commit, is enough to evaluate
if the approach is sounding.
 2019-05-06 18:02:51 +02:00
Salvatore Sanfilippo
179c89c6a4 Merge pull request #5971 from devnexen/unstable 
build fix
 2019-04-26 17:32:21 +02:00
David Carlier
0e00a99f32 build fix 2019-03-28 06:38:16 +00:00
Oran Agra
48d14e5aa7 slave corrupts replication stream when module blocked client uses large reply (or POSTPONED_ARRAY) 
when redis appends the blocked client reply list to the real client, it didn't
bother to check if it is in fact the master client. so a slave executing that
module command will send replies to the master, causing the master to send the
slave error responses, which will mess up the replication offset
(slave will advance it's replication offset, and the master does not)
 2019-03-24 14:17:37 +02:00
Oran Agra
eb40ac6c8e diskless fork kept streaming RDB to a disconnected slave 2019-03-21 20:24:52 +02:00
antirez
2fd9b52b42 Gopher: reply in gopher mode only if argv[0] starts with slash. 
As documented but never implemented.
 2019-02-27 22:20:31 +01:00
antirez
8b01d81fd4 Merge branch 'gopher' into unstable 2019-02-25 18:16:58 +01:00
antirez
43b7a5fa1c RESP3: SETNAME option for HELLO. 2019-02-25 16:56:58 +01:00
antirez
ea75d7d52b RESP3: AUTH option for HELLO. 2019-02-25 16:55:16 +01:00
antirez
17e9ec2a69 RESP3: refactoring of CLIENT SETNAME to implement SETNAME in HELLO. 2019-02-25 16:51:49 +01:00
antirez
505815da4b RESP3: allow HELLO to be used with version = 2. 2019-02-25 16:41:00 +01:00
antirez
dec44d23d3 ACL: move AUTH implementation in acl.c. 2019-02-25 16:33:38 +01:00
antirez
c8b67fea44 Gopher: initial request handling. 2019-02-21 23:13:08 +01:00
Madelyn Olson
79eaa9ebd9 Refactored manual computation of object length 2019-02-21 21:35:00 +00:00
antirez
34f11c811d ACL: when client->user is NULL the client is a superuser. 
Related to #5832.
 2019-02-12 09:44:30 +01:00
zhaozhao.zz
9150d9246e ACL: show client's user 2019-02-12 16:03:58 +08:00
antirez
a5e69e6729 ACL: remove server.requirepass + some refactoring. 2019-01-18 11:49:30 +01:00
antirez
8ea6409e38 ACL: automatically authenticate the nopass default user. 2019-01-15 17:57:49 +01:00
antirez
8df5007346 ACL: avoid a radix tree lookup for the default user. 2019-01-11 11:32:41 +01:00
antirez
33ff2b2f35 ACL: add a reference to the user in each client. 2019-01-10 16:34:13 +01:00
antirez
8cff8ef2d3 ACL: HELLO should stop if the user is not authenticated. 2019-01-09 17:00:30 +01:00
antirez
89b7b6a917 RESP3: addReplyString() -> addReplyProto(). 
The function naming was totally nuts. Let's fix it as we break PRs
anyway with RESP3 refactoring and changes.
 2019-01-09 17:00:30 +01:00
antirez
78a3a19b2c RESP3: verbatim reply API + DEBUG PROTOCOL support. 2019-01-09 17:00:30 +01:00
antirez
6525bd9f68 RESP3: Fix addReplyBool() RESP2/3 output. 2019-01-09 17:00:30 +01:00
antirez
a6ead03510 RESP3: addReplyBool() implemented. 2019-01-09 17:00:29 +01:00
antirez
bf3d7bbf31 RESP3: initial implementation of the HELLO command. 2019-01-09 17:00:29 +01:00
antirez
8ecf7693bf RESP3: addReplyNullArray() added for better RESP2 compat. 2019-01-09 17:00:29 +01:00
antirez
c8304b099d RESP3: most null replies converted. 2019-01-09 17:00:29 +01:00
antirez
29bbe91392 RESP3: addReplyNull() added. 2019-01-09 17:00:29 +01:00
antirez
56c95799a3 RESP3: bring RESP2 compatibility to previous changes. 2019-01-09 17:00:29 +01:00
antirez
e6a467814a RESP3: addReply*Len() support for RESP2 backward comp. 2019-01-09 17:00:29 +01:00
antirez
9325d288f6 RESP3: put RESP version in the client structure. 2019-01-09 17:00:29 +01:00
antirez
689949fb30 RESP3: Aggregate deferred lengths functions. 2019-01-09 17:00:29 +01:00
antirez
7b7a468949 RESP3: Double replies and aggregate lengths initial functions. 2019-01-09 17:00:29 +01:00
antirez
93f443510b Crashing is too much in addReplyErrorLength(). 
See #5663.
 2018-12-11 17:50:18 +01:00
zhaozhao.zz
000eb0fcec networking: current_client should not be NULL when trim qb_pos 2018-12-07 19:14:33 +08:00
Madelyn Olson
e2f8bebfbb Fixed a serverPanic when sending an invalid command to a monitor client 2018-12-04 07:17:17 +00:00
antirez
5789a7ef6d asyncCloseClientOnOutputBufferLimitReached(): don't free fake clients. 
Fake clients are used in special situations and are not linked to the
normal clients list, freeing them will always result in Redis crashing
in one way or the other.

It's not common to send replies to fake clients, but we have one usage
in the modules API. When a client is blocked, we associate to the
blocked client object (that is safe to manipulate in a thread), a fake
client that accumulates replies. So because of this bug there was
the problem described in issue #5443.

The fix was verified to work with the provided example module. To write
a regression is very hard and unlikely to be triggered in the future.
 2018-10-30 13:38:41 +01:00
zhaozhao.zz
138e48f940 Avoid recreate write handler for protected client. 2018-10-09 20:34:11 +08:00