Commit Graph

267 Commits

Author SHA1 Message Date
Erik Johnston
f4f223aa44 Don't make temporary list 2018-10-29 14:01:49 +00:00
Erik Johnston
b2399f6281 Make SQL a bit cleaner 2018-10-29 14:01:11 +00:00
Erik Johnston
03e634dad4 Merge branch 'develop' of github.com:matrix-org/synapse into erikj/purge_state_groups 2018-10-26 16:22:45 +01:00
Erik Johnston
cb53ce9d64
Refactor state group lookup to reduce DB hits (#4011)
Currently when fetching state groups from the data store we make two
hits two the database: once for members and once for non-members (unless
request is filtered to one or the other). This adds needless load to the
datbase, so this PR refactors the lookup to make only a single database
hit.
2018-10-25 17:49:55 +01:00
Erik Johnston
47a9da28ca Batch process handling state groups 2018-10-19 15:48:15 +01:00
Erik Johnston
67a1e315cc Fix up comments 2018-10-12 13:49:48 +01:00
Erik Johnston
4917ff5523 Add state_group index to event_to_state_groups
This is needed to efficiently check for unreferenced state groups during
purge.
2018-10-04 16:03:08 +01:00
Erik Johnston
17d585753f Delete unreferened state groups during purge 2018-10-04 16:03:06 +01:00
Richard van der Hoff
ae6ad4cf41
docstrings and unittests for storage.state (#3958)
I spent ages trying to figure out how I was going mad...
2018-09-27 11:22:25 +01:00
Matthew Hodgson
bb81e78ec6
Split the state_group_cache in two (#3726)
Splits the state_group_cache in two.

One half contains normal state events; the other contains member events.

The idea is that the lazyloading common case of: "I want a subset of member events plus all of the other state" can be accomplished efficiently by splitting the cache into two, and asking for "all events" from the non-members cache, and "just these keys" from the members cache.  This means we can avoid having to make DictionaryCache aware of these sort of complicated queries, whilst letting LL requests benefit from the caching.

Previously we were unable to sensibly use the caching and had to pull all state from the DB irrespective of the filtering, which made things slow.  Hopefully fixes https://github.com/matrix-org/synapse/issues/3720.
2018-08-22 00:56:37 +02:00
Matthew Hodgson
3f543dc021 initial cut at a room summary API (#3574) 2018-08-16 09:46:50 +01:00
Matthew Hodgson
2f78f432c4 speed up /members and add at= and membership params (#3568) 2018-08-15 16:35:22 +01:00
Richard van der Hoff
0ca459ea33 Basic support for room versioning
This is the first tranche of support for room versioning. It includes:
 * setting the default room version in the config file
 * new room_version param on the createRoom API
 * storing the version of newly-created rooms in the m.room.create event
 * fishing the version of existing rooms out of the m.room.create event
2018-08-03 16:08:32 +01:00
Matthew Hodgson
bc7944e6d2 switch missing_types to be a bool 2018-07-25 23:36:31 +01:00
Matthew Hodgson
7d9fb88617 incorporate more review. 2018-07-25 16:33:50 +01:00
Matthew Hodgson
cb5c37a57c handle the edge case for _get_some_state_from_cache where types is [] 2018-07-24 20:34:45 +01:00
Matthew Hodgson
e22700c3dd consider non-filter_type types as wildcards, thus missing from the state-group-cache 2018-07-24 13:59:07 +01:00
Matthew Hodgson
cd241d6bda incorporate more review 2018-07-24 12:39:40 +01:00
Matthew Hodgson
efcdacad7d handle case where types is [] on postgres correctly 2018-07-23 22:41:05 +01:00
Matthew Hodgson
254fb430d1 incorporate review 2018-07-23 19:21:20 +01:00
Matthew Hodgson
2f558300cc fix thinkos; unbreak tests 2018-07-19 19:22:27 +01:00
Matthew Hodgson
bcaec2915a incorporate review 2018-07-19 19:03:50 +01:00
Matthew Hodgson
924eb34d94 add a filtered_types param to limit filtering to specific types 2018-07-19 18:32:02 +01:00
Matthew Hodgson
be3adfc331 merge develop pydoc for _get_state_for_groups 2018-07-19 11:26:04 +01:00
Amber Brown
49af402019 run isort 2018-07-09 16:09:20 +10:00
Richard van der Hoff
43e02c409d Disable partial state group caching for wildcard lookups
When _get_state_for_groups is given a wildcard filter, just do a complete
lookup. Hopefully this will give us the best of both worlds by not filling up
the ram if we only need one or two keys, but also making the cache still work
for the federation reader usecase.
2018-06-22 11:52:07 +01:00
Richard van der Hoff
240f192523
Merge pull request #3382 from matrix-org/rav/optimise_state_groups
Optimise state_group_cache update
2018-06-22 11:20:20 +01:00
Richard van der Hoff
eb32b2ca20 Optimise state_group_cache update
(1) matrix-org-hotfixes has removed the intern calls; let's do the same here.
(2) remove redundant iteritems() so we can used an optimised db update.
2018-06-11 22:56:11 +01:00
Matthew Hodgson
c96d882a02 Merge branch 'develop' into matthew/filter_members 2018-06-10 12:26:14 +03:00
Richard van der Hoff
d3a8c9c55e Fix sql error in _get_state_groups_from_groups
If this was called with a `(type, None)` entry in types (which is supposed to
return all state of type `type`), it would explode with a sql error.
2018-06-06 14:19:01 +01:00
Erik Johnston
042eedfa2b Add hacky cache factor override system 2018-06-04 15:39:28 +01:00
Matthew Hodgson
28f09fcdd5 Merge branch 'develop' into matthew/filter_members 2018-06-04 00:09:17 +03:00
Amber Brown
c936a52a9e
Consistently use six's iteritems and wrap lazy keys/values in list() if they're not meant to be lazy (#3307) 2018-05-31 19:03:47 +10:00
Matthew Hodgson
a6c8f7c875 add pydoc 2018-05-29 01:09:55 +01:00
Matthew Hodgson
9f77001e27 pep8 2018-03-14 00:07:47 +00:00
Matthew Hodgson
d144ed6ffb
fix bug #2926 (loading all state for a given type from the DB if the state_key is None) (#2990)
Fixes a regression that had crept in where the caching layer upholds requests for loading state which is filtered by type (but not by state_key), but the DB layer itself would interpret a missing state_key as a request to filter by null state_key rather than returning all state_keys.
2018-03-13 22:36:04 +00:00
Matthew Hodgson
ccca02846d make it work 2018-03-13 22:31:41 +00:00
Matthew Hodgson
f0f9a0605b remove comment now #2969 is fixed 2018-03-13 22:12:15 +00:00
Matthew Hodgson
12350e3f9a merge proper fix to bug 2969 2018-03-13 22:11:58 +00:00
Matthew Hodgson
afbf4d3dcc typoe 2018-03-13 19:48:04 +00:00
Matthew Hodgson
865377a70d disable optimisation for searching for state groups
when type filter includes wildcards on state_key
2018-03-13 19:46:04 +00:00
Matthew Hodgson
b2aba9e430 build where_clause sanely 2018-03-13 18:13:44 +00:00
Matthew Hodgson
52f7e23c72 PR feedbackz 2018-03-13 18:07:55 +00:00
Matthew Hodgson
1b1c137771 fix bug #2926 2018-03-13 17:52:52 +00:00
Matthew Hodgson
fdedcd1f4d correctly handle None state_keys
and fix include_other_types thinko
2018-03-12 01:39:06 +00:00
Matthew Hodgson
97c0496cfa fix sqlite where clause 2018-03-12 00:27:06 +00:00
Matthew Hodgson
8713365265 typos 2018-03-11 20:10:25 +00:00
Matthew Hodgson
9b334b3f97 WIP experiment in lazyloading room members 2018-03-11 20:01:41 +00:00
Erik Johnston
106906a65e Don't serialize current state over replication 2018-02-15 13:53:18 +00:00
Erik Johnston
fd1601c596 Fix state group storage bug in workers
We needed to move `_count_state_group_hops_txn` to the
StateGroupWorkerStore.
2018-02-15 11:04:32 +00:00
Erik Johnston
3d33eef6fc
Store state groups separately from events (#2784)
* Split state group persist into seperate storage func

* Add per database engine code for state group id gen

* Move store_state_group to StateReadStore

This allows other workers to use it, and so resolve state.

* Hook up store_state_group

* Fix tests

* Rename _store_mult_state_groups_txn

* Rename StateGroupReadStore

* Remove redundant _have_persisted_state_group_txn

* Update comments

* Comment compute_event_context

* Set start val for state_group_id_seq

... otherwise we try to recreate old state groups

* Update comments

* Don't store state for outliers

* Update comment

* Update docstring as state groups are ints
2018-02-06 14:31:24 +00:00
Richard van der Hoff
35a4b63240 Pull out bits of StateStore to a mixin
... so that we don't need to secretly gut-wrench it for use in the slaved
stores. I haven't done the other stores yet, but we should. I'm tired of the
workers breaking every time we tweak the stores because I forgot to gut-wrench
the right method.

fixes https://github.com/matrix-org/synapse/issues/2655.
2017-11-14 11:43:58 +00:00
Richard van der Hoff
4dd1bfa8c1 Revert "Revert "move _state_group_cache to statestore""
We're going to fix this properly on this branch, so that the _state_group_cache
can end up in StateGroupReadStore.

This reverts commit ab335edb02.
2017-11-14 11:43:58 +00:00
Richard van der Hoff
6cfee09be9 Make __init__ consitstent across Store heirarchy
Add db_conn parameters to the `__init__` methods of the *Store classes, so that
they are all consistent, which makes the multiple inheritance work correctly
(and so that we can later extract mixins which can be used in the slavedstores)
2017-11-13 10:46:07 +00:00
Erik Johnston
ab335edb02 Revert "move _state_group_cache to statestore"
This reverts commit f5cf3638e9.
2017-11-13 10:05:33 +00:00
Richard van der Hoff
f5cf3638e9 move _state_group_cache to statestore
this is internal to statestore, so let's keep it there.
2017-11-07 16:43:00 +00:00
Erik Johnston
5946aa0877 Prefill forward extrems and event to state groups 2017-06-29 15:38:48 +01:00
Erik Johnston
8060974344 Fix replication 2017-06-09 16:40:52 +01:00
Erik Johnston
b0d975e216 Comments 2017-06-09 16:25:42 +01:00
Erik Johnston
e54d7d536e Cache state deltas 2017-06-09 16:24:00 +01:00
Erik Johnston
ea11ee09f3 Ensure we don't use unpersisted state group as prev group 2017-06-08 11:59:57 +01:00
Erik Johnston
6ba21bf2b8 Comments 2017-06-07 11:08:36 +01:00
Erik Johnston
dfbda5e025 Faster cache for get_joined_hosts 2017-05-25 17:24:44 +01:00
Erik Johnston
bbfe4e996c Make get_state_groups_from_groups faster.
Most of the time was spent copying a dict to filter out sentinel values
that indicated that keys did not exist in the dict. The sentinel values
were added to ensure that we cached the non-existence of keys.

By updating DictionaryCache to keep track of which keys were known to
not exist itself we can remove a dictionary copy.
2017-05-17 15:12:15 +01:00
Erik Johnston
608b5a6317 Take a copy before prefilling, as it may be a frozendict 2017-05-16 12:55:29 +01:00
Erik Johnston
e4435b014e Update comment 2017-05-15 15:11:30 +01:00
Erik Johnston
871605f4e2 Comments 2017-05-15 15:11:30 +01:00
Erik Johnston
bfbc907cec Prefill state caches 2017-05-15 15:11:13 +01:00
Erik Johnston
587f07543f Revert "Prefill state caches" 2017-05-04 15:07:27 +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
a2c89a225c Prefill state caches 2017-05-02 10:40:31 +01:00
Erik Johnston
f053a1409e Make state caches cache in ascii 2017-04-25 17:22:55 +01:00
Erik Johnston
22f3d3ae76 Reduce _get_state_group_for_event cache size 2017-04-25 11:43:03 +01:00
Erik Johnston
e4f3431116 Remove unused cache 2017-04-24 13:27:38 +01:00
Erik Johnston
2a3e822f44 Comment 2017-04-07 13:47:04 +01:00
Erik Johnston
d72667fcce Speed up get_current_state_ids
Using _simple_select_list is fairly expensive for functions that return
a lot of rows and/or get called a lot. (This is because it carefully
constructs a list of dicts).

get_current_state_ids gets called a lot on startup and e.g. when the IRC
bridge decided to send tonnes of joins/leaves (as it invalidates the
cache). We therefore replace it with a custon txn function that builds
up the final result dict without building up and intermediate
representation.
2017-04-07 10:10:49 +01:00
Erik Johnston
3ce8d59176 Increase cache size for _get_state_group_for_event 2017-03-29 14:31:46 +01:00
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
e71940aa64 Use iter(items|values) 2017-03-24 10:57:02 +00:00
Erik Johnston
00957d1aa4 User Cursor.__iter__ instead of fetchall
This prevents unnecessary construction of lists
2017-03-23 17:53:49 +00:00
Richard van der Hoff
0c01f829ae Avoid resetting state on rejected events
When we get a rejected event, give it the same state_group as its prev_event,
rather than no state_group at all.

This should fix https://github.com/matrix-org/synapse/issues/1935.
2017-03-17 15:06:08 +00:00
Erik Johnston
8ffbe43ba1 Get current state by using current_state_events table 2017-03-10 17:39:35 +00:00
Richard van der Hoff
fc2f29c1d0 Fix bugs in the /keys/changes api
* `get_forward_extremeties_for_room` takes a numeric `stream_ordering`. We were
  passing a `RoomStreamToken`, which meant that it returned the *current*
  extremities, rather than those corresponding to the `from_token`. However:
* `get_state_ids_for_events` required a second ('types') parameter; this meant
  that a `TypeError` was thrown and we ended up acting as though there was *no*
  prev state.
* `get_state_ids_for_events` actually returns a map from event_id to state
  dictionary - just looking up the state keys in it again meant that we acted
  as though there was no prev state. We now check if each member's state has
  changed since *any* of the extremities.

Also add/fix some comments.
2017-02-14 13:59:50 +00:00
Erik Johnston
21b7375778 Add an index to make membership queries faster 2017-01-31 15:15:57 +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
897f8752da Up cache max entries for state 2017-01-16 15:08:17 +00:00
Erik Johnston
01521299c7 Increase cache size limit 2017-01-16 11:56:51 +00:00
Erik Johnston
2fae34bd2c Optionally measure size of cache by sum of length of values 2017-01-13 17:46:17 +00:00
Matthew Hodgson
8cfc0165e9 fix annoying typos 2017-01-05 13:39:43 +00:00
Erik Johnston
587d8ac60f Correctly intern keys in state cache 2016-11-08 11:53:25 +00:00
Erik Johnston
4974147aa3 Remove duplication 2016-09-27 09:27:54 +01:00
Erik Johnston
13122e5e24 Remove unused variable 2016-09-27 09:21:51 +01:00
Erik Johnston
cf3e1cc200 Fix perf of fetching state in SQLite 2016-09-26 17:16:24 +01:00
Erik Johnston
00f51493f5 Fix reindex 2016-09-14 10:18:30 +01:00
Erik Johnston
d5ae1f1291 Ensure we don't mutate state cache entries 2016-09-14 10:03:48 +01:00
Erik Johnston
03a98aff3c Create new index concurrently 2016-09-12 14:27:01 +01:00
Erik Johnston
897d57bc58 Change state fetch query for postgres to be faster
It turns out that postgres doesn't like doing a list of OR's and is
about 1000x slower, so we just issue a query for each specific type
seperately.
2016-09-12 10:05:07 +01:00
Erik Johnston
5beda10bbd Reindex state_groups_state after pruning 2016-09-08 16:18:01 +01:00
Erik Johnston
b568ca309c Temporarily disable sequential scans for state fetching 2016-09-08 09:38:54 +01:00