Commit Graph

1991 Commits

Author SHA1 Message Date
Richard van der Hoff
ff3d810ea8 Add a comment to old delta 2017-05-11 12:48:50 +01:00
Richard van der Hoff
34194aaff7 Don't create event_search index on sqlite
... because the table is virtual
2017-05-11 12:46:55 +01:00
Richard van der Hoff
114f290947 Add more logging for purging
Log the number of events we will be deleting at info.
2017-05-11 12:08:47 +01:00
Richard van der Hoff
baafb85ba4 Add an index to event_search
- to make the purge API quicker
2017-05-11 12:05:22 +01:00
Richard van der Hoff
dc026bb16f Tidy purge code and add some comments
Try to make this clearer with more comments and some variable renames
2017-05-11 10:56:12 +01:00
Richard van der Hoff
8e345ce465 Don't de-delta state groups we're about to delete 2017-05-10 18:44:22 +01:00
Richard van der Hoff
b64d312421 add some logging to purge_history 2017-05-10 18:44:22 +01:00
Erik Johnston
a3648f84b2 Merge pull request #2208 from matrix-org/erikj/ratelimit_overrid
Add per user ratelimiting overrides
2017-05-10 15:54:48 +01:00
Erik Johnston
b990b2fce5 Add per user ratelimiting overrides 2017-05-10 11:05:43 +01:00
Richard van der Hoff
a7e9d8762d Allow clients to upload one-time-keys with new sigs
When a client retries a key upload, don't give an error if the signature has
changed (but the key is the same).

Fixes https://github.com/vector-im/riot-android/issues/1208, hopefully.
2017-05-09 19:04:56 +01:00
Erik Johnston
40dcf0d856 Merge pull request #2203 from matrix-org/erikj/event_cache_hit_ratio
Don't update event cache hit ratio from get_joined_users
2017-05-08 16:52:39 +01:00
Erik Johnston
d3c3026496 Merge pull request #2201 from matrix-org/erikj/store_device_cache
Cache check to see if device exists
2017-05-08 16:23:04 +01:00
Erik Johnston
093f7e47cc Expand docstring a bit 2017-05-08 16:14:46 +01:00
Erik Johnston
6a12998a83 Add missing yields 2017-05-08 16:10:51 +01:00
Erik Johnston
ffad4fe35b Don't update event cache hit ratio from get_joined_users
Otherwise the hit ration of plain get_events gets completely skewed by
calls to get_joined_users* functions.
2017-05-08 16:06:17 +01:00
Erik Johnston
94e6ad71f5 Invalidate cache on device deletion 2017-05-08 15:55:59 +01:00
Erik Johnston
8571f864d2 Cache one time key counts 2017-05-08 15:34:27 +01:00
Erik Johnston
fc6d4974a6 Comment 2017-05-08 15:33:57 +01:00
Erik Johnston
738ccf61c0 Cache check to see if device exists 2017-05-08 15:32:18 +01:00
Erik Johnston
dcabef952c Increase client_ip cache size 2017-05-08 15:09:19 +01:00
Erik Johnston
e0f20e9425 Revert "Remove unused import"
This reverts commit ab37bef83b.
2017-05-08 13:07:43 +01:00
Erik Johnston
fe7c1b969c Revert "We don't care about forgotten rooms"
This reverts commit ad8b316939.
2017-05-08 13:07:43 +01:00
Erik Johnston
78f306a6f7 Revert "Speed up filtering of a single event in push"
This reverts commit 421fdf7460.
2017-05-08 13:07:41 +01:00
Erik Johnston
66f371e8b8 Merge pull request #2176 from matrix-org/erikj/faster_get_joined
Make get_joined_users faster
2017-05-05 10:59:55 +01:00
Erik Johnston
e3aa8a7aa8 Merge pull request #2185 from matrix-org/erikj/smaller_caches
Optimise caches for single key
2017-05-05 10:19:05 +01:00
Erik Johnston
cf589f2c1e Fixes 2017-05-05 10:17:56 +01:00
Erik Johnston
8af4569583 Merge pull request #2174 from matrix-org/erikj/current_cache_hosts
Add cache for get_current_hosts_in_room
2017-05-05 10:15:24 +01:00
Erik Johnston
587f07543f Revert "Prefill state caches" 2017-05-04 15:07:27 +01:00
Erik Johnston
aa93cb9f44 Add comment 2017-05-04 14:59:28 +01:00
Erik Johnston
537dbadea0 Intern host strings 2017-05-04 14:55:28 +01:00
Erik Johnston
07a07588a0 Make caches bigger 2017-05-04 14:52:28 +01:00
Erik Johnston
dfaa58f72d Fix comment and num args 2017-05-04 14:50:24 +01:00
Erik Johnston
5d8290429c Reduce size of get_users_in_room 2017-05-04 13:43:19 +01:00
Erik Johnston
7ebf518c02 Make get_joined_users faster 2017-05-03 15:55:54 +01:00
Erik Johnston
2c2dcf81d0 Update comment 2017-05-03 10:00:29 +01:00
Erik Johnston
1827057acc Comments 2017-05-03 09:56:05 +01:00
Erik Johnston
8346e6e696 Merge branch 'develop' of github.com:matrix-org/synapse into erikj/prefill_state 2017-05-03 09:46:40 +01:00
Erik Johnston
3e5a62ecd8 Add more granular event send metrics 2017-05-02 14:23:26 +01:00
Erik Johnston
a2c89a225c Prefill state caches 2017-05-02 10:40:31 +01:00
Erik Johnston
7166854f41 Add cache for get_current_hosts_in_room 2017-05-02 10:36:35 +01:00
Erik Johnston
3033261891 Merge pull request #2080 from matrix-org/erikj/filter_speed
Speed up filtering of a single event in push
2017-04-28 14:17:13 +01:00
Erik Johnston
ab37bef83b Remove unused import 2017-04-28 09:57:23 +01:00
Erik Johnston
ad8b316939 We don't care about forgotten rooms 2017-04-28 09:52:36 +01:00
Erik Johnston
421fdf7460 Speed up filtering of a single event in push 2017-04-28 09:52:36 +01:00
Richard van der Hoff
c84770b877 Fix bgupdate error if index already exists (#2167)
When creating a new table index in the background, guard against it existing already. Fixes
https://github.com/matrix-org/synapse/issues/2135.

Also, make sure we restore the autocommit flag when we're done, otherwise we
get more failures from other operations later on. Fixes
https://github.com/matrix-org/synapse/issues/1890 (hopefully).
2017-04-27 15:27:48 +01:00
Erik Johnston
f053a1409e Make state caches cache in ascii 2017-04-25 17:22:55 +01:00
Erik Johnston
f144365281 Comment 2017-04-25 15:18:26 +01:00
Erik Johnston
d9aa645f86 Reduce size of joined_user cache
The _get_joined_users_from_context cache stores a mapping from user_id
to avatar_url and display_name. Instead of storing those in a dict,
store them in a namedtuple as that uses much less memory.

We also try converting the string to ascii to further reduce the size.
2017-04-25 14:38:51 +01:00
Erik Johnston
22f3d3ae76 Reduce _get_state_group_for_event cache size 2017-04-25 11:43:03 +01:00
Erik Johnston
119cb9bbcf Reduce cache size by not storing deferreds
Currently the cache descriptors store deferreds rather than raw values,
this is a simple way of triggering only one database hit and sharing the
result if two callers attempt to get the same value.

However, there are a few caches that simply store a mapping from string
to string (or int). These caches can have a large number of entries,
under the assumption that each entry is small. However, the size of a
deferred (specifically the size of ObservableDeferred) is signigicantly
larger than that of the raw value, 2kb vs 32b.

This PR therefore changes the cache descriptors to store the raw values
rather than the deferreds.

As a side effect cached storage function now either return a deferred or
the actual value, as the cached list decriptor already does. This is
fine as we always end up just yield'ing on the returned value
eventually, which handles that case correctly.
2017-04-25 10:23:11 +01:00