Commit Graph

129 Commits

Author SHA1 Message Date
Erik Johnston
ab904caf33 Comments 2017-04-10 10:02:17 +01:00
Erik Johnston
98ce212093 Merge pull request #2103 from matrix-org/erikj/no-double-encode
Don't double encode replication data
2017-04-07 09:39:52 +01:00
Erik Johnston
ad544c803a Document types of the replication streams 2017-04-06 13:28:52 +01:00
Erik Johnston
69b3fd485d Fix incorrect type when using InvalidateCacheCommand 2017-04-06 09:36:38 +01:00
Erik Johnston
fcc803b2bf Add log lines 2017-04-05 17:13:44 +01:00
Erik Johnston
3f213d908d Rearrange metrics 2017-04-05 14:15:09 +01:00
Erik Johnston
1ca0e78ca1 Fix typo 2017-04-05 13:43:39 +01:00
Erik Johnston
b43d3267e2 Fixup some metrics for tcp repl 2017-04-05 13:34:54 +01:00
Erik Johnston
a5c401bd12 Merge pull request #2097 from matrix-org/erikj/repl_tcp_client
Move to using TCP replication
2017-04-05 09:36:21 +01:00
Erik Johnston
a76886726b Merge pull request #2098 from matrix-org/erikj/repl_tcp_fix
Advance replication streams even if nothing is listening
2017-04-04 15:40:51 +01:00
Erik Johnston
4264ceb31c Fiddle tcp replication logging 2017-04-04 14:14:03 +01:00
Erik Johnston
023ee197be Advance replication streams even if nothing is listening
Otherwise the streams don't advance and steadily fall behind, so when a
worker does connect either a) they'll be streamed lots of old updates or
b) the connection will fail as the streams are too far behind.
2017-04-04 13:19:26 +01:00
Erik Johnston
3a1f3f8388 Change slave storage to use new replication interface
As the TCP replication uses a slightly different API and streams than
the HTTP replication.

This breaks HTTP replication.
2017-04-03 15:34:19 +01:00
Erik Johnston
52bfa604e1 Add basic replication client handler and factory 2017-04-03 15:34:13 +01:00
Erik Johnston
0a6a966e2b Always advance stream tokens 2017-04-03 15:22:56 +01:00
Erik Johnston
1df7c28661 Use callbacks to notify tcp replication rather than deferreds 2017-03-31 15:42:51 +01:00
Erik Johnston
36d2b66f90 Add a timestamp to USER_SYNC command
This timestamp is used to indicate when the user last sync'd
2017-03-31 15:42:22 +01:00
Erik Johnston
bfcf016714 Fix up docs 2017-03-31 11:19:24 +01:00
Erik Johnston
4d7fc7f977 Add server side resource for tcp replication 2017-03-30 13:24:45 +01:00
Erik Johnston
7450693435 Initial TCP protocol implementation
This defines the low level TCP replication protocol
2017-03-30 12:54:46 +01:00
Erik Johnston
8da6f0be48 Define the various streams we will replicate 2017-03-30 12:54:46 +01:00
Erik Johnston
11880103b1 Make federation send queue take the current position 2017-03-30 12:54:36 +01:00
Erik Johnston
24d35ab47b Add new storage functions for new replication
The new replication protocol will keep all the streams separate, rather
than muxing multiple streams into one.
2017-03-30 11:48:35 +01:00
Erik Johnston
09f79aaad0 Use presence replication stream to invalidate cache
Instead of using the cache invalidation replication stream to invalidate
the _get_presence_cache, we can instead rely on the presence replication
stream. This reduces the amount of replication traffic considerably.
2017-03-24 13:21:08 +00:00
Erik Johnston
d58b1ffe94 Replace some calls to cursor_to_dict
cursor_to_dict can be surprisinglh expensive for large result sets, so lets
only call it when we need to.
2017-03-24 11:07:02 +00:00
Erik Johnston
aac6d1fc9b PEP8 2017-03-20 13:47:56 +00:00
Erik Johnston
61f471f779 Don't send the full event json over replication 2017-03-17 15:50:01 +00:00
Richard van der Hoff
29ed09e80a Fix assertion to stop transaction queue getting wedged
... and update some docstrings to correctly reflect the types being used.

get_new_device_msgs_for_remote can return a long under some circumstances,
which was being stored in last_device_list_stream_id_by_dest, and was then
upsetting things on the next loop.
2017-03-15 12:16:55 +00:00
Erik Johnston
45c7f12d2a Add new storage function to slave store 2017-03-13 16:26:44 +00:00
Erik Johnston
8f267fa8a8 Fix it for the workers 2017-03-10 11:22:25 +00:00
Erik Johnston
e933a2712d Don't log unknown cache warnings in workers 2017-02-28 16:22:41 +00:00
Erik Johnston
095b45c165 Aggregate event push actions 2017-02-14 13:39:41 +00:00
Erik Johnston
9e617cd4c2 Cache get_presence storage 2017-02-13 13:50:03 +00:00
Erik Johnston
6bba80241c Merge pull request #1912 from matrix-org/markjh/roominitialsync
Add db functions needed for room initial sync to slave
2017-02-13 12:20:21 +01:00
Mark Haines
3a46280ca3 Add db functions needed for room initial sync to slave 2017-02-13 11:16:53 +00:00
Erik Johnston
0f3e296cb7 Fix replication 2017-02-02 15:02:03 +00:00
Erik Johnston
458b6f4733 Only invalidate membership caches based on the cache stream
Before we completely invalidated get_users_in_room whenever we updated
any current_state_events table. This was way too aggressive.
2017-01-31 16:09:03 +00:00
Erik Johnston
3670025e64 Rename func 2017-01-30 14:11:31 +00:00
Erik Johnston
252b503fc8 Hook device list updates to replication 2017-01-27 14:31:35 +00:00
Erik Johnston
a55fa2047f Insert delta of current_state_events to be more efficient 2017-01-20 17:10:18 +00:00
Erik Johnston
09cbcb78d3 Add cache to get_public_room_ids_at_stream_id 2016-12-12 14:41:51 +00:00
Erik Johnston
f32fb65552 Add new API appservice specific public room list 2016-12-06 16:12:27 +00:00
Erik Johnston
26072df6af Ensure only main or federation_sender process can send federation traffic 2016-11-23 14:09:47 +00:00
Erik Johnston
4c79a63fd7 Explicit federation ack 2016-11-23 10:40:44 +00:00
Erik Johnston
90565d015e Invalidate retry cache in both directions 2016-11-22 17:45:44 +00:00
Erik Johnston
7c9cdb2245 Store federation stream positions in the database 2016-11-21 11:33:08 +00:00
Erik Johnston
f8ee66250a Handle sending events and device messages over federation 2016-11-17 15:48:04 +00:00
Erik Johnston
ed787cf09e Hook up the send queue and create a federation sender worker 2016-11-16 17:34:44 +00:00
Erik Johnston
668f91d707 Fix check of wrong variable 2016-10-11 13:57:22 +01:00
Erik Johnston
748d8fdc7b Reduce DB hits for replication
Some streams will occaisonally advance their positions without actually
having any new rows to send over federation. Currently this means that
the token will not advance on the workers, leading to them repeatedly
sending a slightly out of date token. This in turns requires the master
to hit the DB to check if there are any new rows, rather than hitting
the no op logic where we check if the given token matches the current
token.

This commit changes the API to always return an entry if the position
for a stream has changed, allowing workers to advance their tokens
correctly.
2016-09-23 16:49:21 +01:00