Commit Graph

780 Commits

Author SHA1 Message Date
reivilibre
c06b2b7142
Faster Remote Room Joins: tell remote homeservers that we are unable to authorise them if they query a room which has partial state on our server. (#13823) 2022-09-23 11:47:16 +01:00
Eric Eastwood
51d732db3b
Optimize how we calculate likely_domains during backfill (#13575)
Optimize how we calculate `likely_domains` during backfill because I've seen this take 17s in production just to `get_current_state` which is used to `get_domains_from_state` (see case [*2. Loading tons of events* in the `/messages` investigation issue](https://github.com/matrix-org/synapse/issues/13356)).

There are 3 ways we currently calculate hosts that are in the room:

 1. `get_current_state` -> `get_domains_from_state`
    - Used in `backfill` to calculate `likely_domains` and `/timestamp_to_event` because it was cargo-culted from `backfill`
    - This one is being eliminated in favor of `get_current_hosts_in_room` in this PR 🕳
 1. `get_current_hosts_in_room`
    - Used for other federation things like sending read receipts and typing indicators
 1. `get_hosts_in_room_at_events`
    - Used when pushing out events over federation to other servers in the `_process_event_queue_loop`

Fix https://github.com/matrix-org/synapse/issues/13626

Part of https://github.com/matrix-org/synapse/issues/13356

Mentioned in [internal doc](https://docs.google.com/document/d/1lvUoVfYUiy6UaHB6Rb4HicjaJAU40-APue9Q4vzuW3c/edit#bookmark=id.2tvwz3yhcafh)


### Query performance

#### Before

The query from `get_current_state` sucks just because we have to get all 80k events. And we see almost the exact same performance locally trying to get all of these events (16s vs 17s):
```
synapse=# SELECT type, state_key, event_id FROM current_state_events WHERE room_id = '!OGEhHVWSdvArJzumhm:matrix.org';
Time: 16035.612 ms (00:16.036)

synapse=# SELECT type, state_key, event_id FROM current_state_events WHERE room_id = '!OGEhHVWSdvArJzumhm:matrix.org';
Time: 4243.237 ms (00:04.243)
```

But what about `get_current_hosts_in_room`: When there is 8M rows in the `current_state_events` table, the previous query in `get_current_hosts_in_room` took 13s from complete freshness (when the events were first added). But takes 930ms after a Postgres restart or 390ms if running back to back to back.

```sh
$ psql synapse
synapse=# \timing on
synapse=# SELECT COUNT(DISTINCT substring(state_key FROM '@[^:]*:(.*)$'))
FROM current_state_events
WHERE
    type = 'm.room.member'
    AND membership = 'join'
    AND room_id = '!OGEhHVWSdvArJzumhm:matrix.org';
 count
-------
  4130
(1 row)

Time: 13181.598 ms (00:13.182)

synapse=# SELECT COUNT(*) from current_state_events where room_id = '!OGEhHVWSdvArJzumhm:matrix.org';
 count
-------
 80814

synapse=# SELECT COUNT(*) from current_state_events;
  count
---------
 8162847

synapse=# SELECT pg_size_pretty( pg_total_relation_size('current_state_events') );
 pg_size_pretty
----------------
 4702 MB
```

#### After

I'm not sure how long it takes from complete freshness as I only really get that opportunity once (maybe restarting computer but that's cumbersome) and it's not really relevant to normal operating times. Maybe you get closer to the fresh times the more access variability there is so that Postgres caches aren't as exact. Update: The longest I've seen this run for is 6.4s and 4.5s after a computer restart.

After a Postgres restart, it takes 330ms and running back to back takes 260ms.

```sh
$ psql synapse
synapse=# \timing on
Timing is on.
synapse=# SELECT
    substring(c.state_key FROM '@[^:]*:(.*)$') as host
FROM current_state_events c
/* Get the depth of the event from the events table */
INNER JOIN events AS e USING (event_id)
WHERE
    c.type = 'm.room.member'
    AND c.membership = 'join'
    AND c.room_id = '!OGEhHVWSdvArJzumhm:matrix.org'
GROUP BY host
ORDER BY min(e.depth) ASC;
Time: 333.800 ms
```

#### Going further

To improve things further we could add a `limit` parameter to `get_current_hosts_in_room`. Realistically, we don't need 4k domains to choose from because there is no way we're going to query that many before we a) probably get an answer or b) we give up. 

Another thing we can do is optimize the query to use a index skip scan:

 - https://wiki.postgresql.org/wiki/Loose_indexscan
 - Index Skip Scan, https://commitfest.postgresql.org/37/1741/
 - https://www.timescale.com/blog/how-we-made-distinct-queries-up-to-8000x-faster-on-postgresql/
2022-08-30 01:38:14 -05:00
Eric Eastwood
9385c41ba4
Fix Prometheus metrics being negative (mixed up start/end) (#13584)
Fix:

 - https://github.com/matrix-org/synapse/pull/13535#discussion_r949582508
 - https://github.com/matrix-org/synapse/pull/13533#discussion_r949577244
2022-08-23 08:47:30 +01:00
Eric Eastwood
088bcb7ecb
Time how long it takes us to do backfill processing (#13535) 2022-08-17 10:33:19 +01:00
Eric Eastwood
0a4efbc1dd
Instrument the federation/backfill part of /messages (#13489)
Instrument the federation/backfill part of `/messages` so it's easier to follow what's going on in Jaeger when viewing a trace.

Split out from https://github.com/matrix-org/synapse/pull/13440

Follow-up from https://github.com/matrix-org/synapse/pull/13368

Part of https://github.com/matrix-org/synapse/issues/13356
2022-08-16 12:39:40 -05:00
Eric Eastwood
344a2f767c
Instrument FederationStateIdsServlet - /state_ids (#13499)
Instrument FederationStateIdsServlet - `/state_ids` so it's easier to follow what's going on in Jaeger when viewing a trace.
2022-08-15 19:41:23 +01:00
reivilibre
e9e6aacfbe
Faster Room Joins: prevent Synapse from answering federated join requests for a room which it has not fully joined yet. (#13416) 2022-08-04 16:27:04 +01:00
Eric Eastwood
92d21faf12
Instrument /messages for understandable traces in Jaeger (#13368)
In Jaeger:

 - Before: huge list of uncategorized database calls
 - After: nice and collapsible into units of work
2022-08-03 10:57:38 -05:00
Sean Quah
8d317f6da5
Fix error when out of servers to sync partial state with (#13432)
so that we raise the intended error instead.

Signed-off-by: Sean Quah <seanq@matrix.org>
2022-08-02 12:12:44 +01:00
reivilibre
e17e5c97e0
Faster Room Joins: don't leave a stuck room partial state flag if the join fails. (#13403) 2022-08-01 16:45:39 +00:00
David Teller
11f811470f
Uniformize spam-checker API, part 5: expand other spam-checker callbacks to return Tuple[Codes, dict] (#13044)
Signed-off-by: David Teller <davidt@element.io>
Co-authored-by: Brendan Abolivier <babolivier@matrix.org>
2022-07-11 16:52:10 +00:00
Sean Quah
1391a76cd2
Faster room joins: fix race in recalculation of current room state (#13151)
Bounce recalculation of current state to the correct event persister and
move recalculation of current state into the event persistence queue, to
avoid concurrent updates to a room's current state.

Also give recalculation of a room's current state a real stream
ordering.

Signed-off-by: Sean Quah <seanq@matrix.org>
2022-07-07 12:19:31 +00:00
Sean Quah
68db233f0c
Handle race between persisting an event and un-partial stating a room (#13100)
Whenever we want to persist an event, we first compute an event context,
which includes the state at the event and a flag indicating whether the
state is partial. After a lot of processing, we finally try to store the
event in the database, which can fail for partial state events when the
containing room has been un-partial stated in the meantime.

We detect the race as a foreign key constraint failure in the data store
layer and turn it into a special `PartialStateConflictError` exception,
which makes its way up to the method in which we computed the event
context.

To make things difficult, the exception needs to cross a replication
request: `/fed_send_events` for events coming over federation and
`/send_event` for events from clients. We transport the
`PartialStateConflictError` as a `409 Conflict` over replication and
turn `409`s back into `PartialStateConflictError`s on the worker making
the request.

All client events go through
`EventCreationHandler.handle_new_client_event`, which is called in
*a lot* of places. Instead of trying to update all the code which
creates client events, we turn the `PartialStateConflictError` into a
`429 Too Many Requests` in
`EventCreationHandler.handle_new_client_event` and hope that clients
take it as a hint to retry their request.

On the federation event side, there are 7 places which compute event
contexts. 4 of them use outlier event contexts:
`FederationEventHandler._auth_and_persist_outliers_inner`,
`FederationHandler.do_knock`, `FederationHandler.on_invite_request` and
`FederationHandler.do_remotely_reject_invite`. These events won't have
the partial state flag, so we do not need to do anything for then.

The remaining 3 paths which create events are
`FederationEventHandler.process_remote_join`,
`FederationEventHandler.on_send_membership_event` and
`FederationEventHandler._process_received_pdu`.

We can't experience the race in `process_remote_join`, unless we're
handling an additional join into a partial state room, which currently
blocks, so we make no attempt to handle it correctly.

`on_send_membership_event` is only called by
`FederationServer._on_send_membership_event`, so we catch the
`PartialStateConflictError` there and retry just once.

`_process_received_pdu` is called by `on_receive_pdu` for incoming
events and `_process_pulled_event` for backfill. The latter should never
try to persist partial state events, so we ignore it. We catch the
`PartialStateConflictError` in `on_receive_pdu` and retry just once.

Refering to the graph of code paths in
https://github.com/matrix-org/synapse/issues/12988#issuecomment-1156857648
may make the above make more sense.

Signed-off-by: Sean Quah <seanq@matrix.org>
2022-07-05 16:12:52 +01:00
David Teller
a164a46038
Uniformize spam-checker API, part 4: port other spam-checker callbacks to return Union[Allow, Codes]. (#12857)
Co-authored-by: Brendan Abolivier <babolivier@matrix.org>
2022-06-13 18:16:16 +00:00
Richard van der Hoff
f68b5e5773 Merge branch 'rav/simplify_event_auth_interface' into develop 2022-06-13 11:34:59 +01:00
Richard van der Hoff
c1b28b8842 Remove redundant room_version param from check_auth_rules_from_context
It's now implied by the room_version property on the event.
2022-06-12 23:13:10 +01:00
Richard van der Hoff
68be42f6b6 Remove room_version param from validate_event_for_room_version
Instead, use the `room_version` property of the event we're validating.

The `room_version` was originally added as a parameter somewhere around #4482,
but really it's been redundant since #6875 added a `room_version` field to `EventBase`.
2022-06-12 23:13:09 +01:00
Richard van der Hoff
7c6b2204d1
Faster joins: add issue links to the TODOs (#13004)
... to help us keep track of these things
2022-06-09 10:13:03 +00:00
Erik Johnston
e3163e2e11
Reduce the amount of state we pull from the DB (#12811) 2022-06-06 09:24:12 +01:00
Erik Johnston
888a29f412
Wait for lazy join to complete when getting current state (#12872) 2022-06-01 16:02:53 +01:00
Sean Quah
641908f72f
Faster room joins: Resume state re-syncing after a Synapse restart (#12813)
Signed-off-by: Sean Quah <seanq@matrix.org>
2022-05-31 15:15:08 +00:00
Sean Quah
2fba1076c5
Faster room joins: Try other destinations when resyncing the state of a partial-state room (#12812)
Signed-off-by: Sean Quah <seanq@matrix.org>
2022-05-31 15:50:29 +01:00
Erik Johnston
1e453053cb
Rename storage classes (#12913) 2022-05-31 12:17:50 +00:00
Erik Johnston
4660d9fdcf
Fix up state_store naming (#12871) 2022-05-25 12:59:04 +01:00
Shay
71e8afe34d
Update EventContext get_current_event_ids and get_prev_event_ids to accept state filters and update calls where possible (#12791) 2022-05-20 09:54:12 +01:00
Erik Johnston
c72d26c1e1
Refactor EventContext (#12689)
Refactor how the `EventContext` class works, with the intention of reducing the amount of state we fetch from the DB during event processing.

The idea here is to get rid of the cached `current_state_ids` and `prev_state_ids` that live in the `EventContext`, and instead defer straight to the database (and its caching). 

One change that may have a noticeable effect is that we now no longer prefill the `get_current_state_ids` cache on a state change. However, that query is relatively light, since its just a case of reading a table from the DB (unlike fetching state at an event which is more heavyweight). For deployments with workers this cache isn't even used.


Part of #12684
2022-05-10 19:43:13 +00:00
andrew do
01e625513a
remove constantly lib use and switch to enums. (#12624) 2022-05-04 11:26:11 +00:00
Richard van der Hoff
17d99f758a
Optimise backfill calculation (#12522)
Try to avoid an OOM by checking fewer extremities.

Generally this is a big rewrite of _maybe_backfill, to try and fix some of the TODOs and other problems in it. It's best reviewed commit-by-commit.
2022-04-26 10:27:11 +01:00
Richard van der Hoff
320186319a
Resync state after partial-state join (#12394)
We work through all the events with partial state, updating the state at each
of them. Once it's done, we recalculate the state for the whole room, and then
mark the room as having complete state.
2022-04-12 13:23:43 +00:00
Sean Quah
800ba87cc8
Refactor and convert Linearizer to async (#12357)
Refactor and convert `Linearizer` to async. This makes a `Linearizer`
cancellation bug easier to fix.

Also refactor to use an async context manager, which eliminates an
unlikely footgun where code that doesn't immediately use the context
manager could forget to release the lock.

Signed-off-by: Sean Quah <seanq@element.io>
2022-04-05 15:43:52 +01:00
Richard van der Hoff
afa17f0eab
Return a 404 from /state for an outlier (#12087)
* Replace `get_state_for_pdu` with  `get_state_ids_for_pdu` and `get_events_as_list`.
* Return a 404 from `/state` and `/state_ids` for an outlier
2022-03-21 11:23:32 +00:00
Richard van der Hoff
dc8d825ef2
Skip attempt to get state at backwards-extremities (#12173)
We don't *have* the state at a backwards-extremity, so this is never going to
do anything useful.
2022-03-09 11:00:48 +00:00
Richard van der Hoff
e2e1d90a5e
Faster joins: persist to database (#12012)
When we get a partial_state response from send_join, store information in the
database about it:
 * store a record about the room as a whole having partial state, and stash the
   list of member servers too.
 * flag the join event itself as having partial state
 * also, for any new events whose prev-events are partial-stated, note that
   they will *also* be partial-stated.

We don't yet make any attempt to interpret this data, so API calls (and a bunch
of other things) are just going to get incorrect data.
2022-03-01 12:49:54 +00:00
Richard van der Hoff
e24ff8ebe3
Remove HomeServer.get_datastore() (#12031)
The presence of this method was confusing, and mostly present for backwards
compatibility. Let's get rid of it.

Part of #11733
2022-02-23 11:04:02 +00:00
Richard van der Hoff
81364db49b
Run _handle_queued_pdus as a background process (#12041)
... to ensure it gets a proper log context, mostly.
2022-02-22 13:33:22 +00:00
Richard van der Hoff
3070af4809
remote join processing: get create event from state, not auth_chain (#12039)
A follow-up to #12005, in which I apparently missed that there are a bunch of other places that assume the create event is in the auth chain.
2022-02-21 19:27:35 +00:00
Eric Eastwood
fef2e792be
Fix historical messages backfilling in random order on remote homeservers (MSC2716) (#11114)
Fix https://github.com/matrix-org/synapse/issues/11091
Fix https://github.com/matrix-org/synapse/issues/10764 (side-stepping the issue because we no longer have to deal with `fake_prev_event_id`)

 1. Made the `/backfill` response return messages in `(depth, stream_ordering)` order (previously only sorted by `depth`)
    - Technically, it shouldn't really matter how `/backfill` returns things but I'm just trying to make the `stream_ordering` a little more consistent from the origin to the remote homeservers in order to get the order of messages from `/messages` consistent ([sorted by `(topological_ordering, stream_ordering)`](https://github.com/matrix-org/synapse/blob/develop/docs/development/room-dag-concepts.md#depth-and-stream-ordering)).
    - Even now that we return backfilled messages in order, it still doesn't guarantee the same `stream_ordering` (and more importantly the [`/messages` order](https://github.com/matrix-org/synapse/blob/develop/docs/development/room-dag-concepts.md#depth-and-stream-ordering)) on the other server. For example, if a room has a bunch of history imported and someone visits a permalink to a historical message back in time, their homeserver will skip over the historical messages in between and insert the permalink as the next message in the `stream_order` and totally throw off the sort.
       - This will be even more the case when we add the [MSC3030 jump to date API endpoint](https://github.com/matrix-org/matrix-doc/pull/3030) so the static archives can navigate and jump to a certain date.
       - We're solving this in the future by switching to [online topological ordering](https://github.com/matrix-org/gomatrixserverlib/issues/187) and [chunking](https://github.com/matrix-org/synapse/issues/3785) which by its nature will apply retroactively to fix any inconsistencies introduced by people permalinking
 2. As we're navigating `prev_events` to return in `/backfill`, we order by `depth` first (newest -> oldest) and now also tie-break based on the `stream_ordering` (newest -> oldest). This is technically important because MSC2716 inserts a bunch of historical messages at the same `depth` so it's best to be prescriptive about which ones we should process first. In reality, I think the code already looped over the historical messages as expected because the database is already in order.
 3. Making the historical state chain and historical event chain float on their own by having no `prev_events` instead of a fake `prev_event` which caused backfill to get clogged with an unresolvable event. Fixes https://github.com/matrix-org/synapse/issues/11091 and https://github.com/matrix-org/synapse/issues/10764
 4. We no longer find connected insertion events by finding a potential `prev_event` connection to the current event we're iterating over. We now solely rely on marker events which when processed, add the insertion event as an extremity and the federating homeserver can ask about it when time calls.
    - Related discussion, https://github.com/matrix-org/synapse/pull/11114#discussion_r741514793


Before | After
--- | ---
![](https://user-images.githubusercontent.com/558581/139218681-b465c862-5c49-4702-a59e-466733b0cf45.png) | ![](https://user-images.githubusercontent.com/558581/146453159-a1609e0a-8324-439d-ae44-e4bce43ac6d1.png)



#### Why aren't we sorting topologically when receiving backfill events?

> The main reason we're going to opt to not sort topologically when receiving backfill events is because it's probably best to do whatever is easiest to make it just work. People will probably have opinions once they look at [MSC2716](https://github.com/matrix-org/matrix-doc/pull/2716) which could change whatever implementation anyway.
> 
> As mentioned, ideally we would do this but code necessary to make the fake edges but it gets confusing and gives an impression of “just whyyyy” (feels icky). This problem also dissolves with online topological ordering.
>
> -- https://github.com/matrix-org/synapse/pull/11114#discussion_r741517138

See https://github.com/matrix-org/synapse/pull/11114#discussion_r739610091 for the technical difficulties
2022-02-07 15:54:13 -06:00
Richard van der Hoff
251b5567ec
Remove log_function and its uses (#11761)
I've never found this terribly useful. I think it was added in the early days
of Synapse, without much thought as to what would actually be useful to log,
and has just been cargo-culted ever since.

Rather, it tends to clutter up debug logs with useless information.
2022-01-18 13:06:04 +00:00
Sean Quah
0147b3de20
Add missing type hints to synapse.logging.context (#11556) 2021-12-14 17:35:28 +00:00
Eric Eastwood
a6f1a3abec
Add MSC3030 experimental client and federation API endpoints to get the closest event to a given timestamp (#9445)
MSC3030: https://github.com/matrix-org/matrix-doc/pull/3030

Client API endpoint. This will also go and fetch from the federation API endpoint if unable to find an event locally or we found an extremity with possibly a closer event we don't know about.
```
GET /_matrix/client/unstable/org.matrix.msc3030/rooms/<roomID>/timestamp_to_event?ts=<timestamp>&dir=<direction>
{
    "event_id": ...
    "origin_server_ts": ...
}
```

Federation API endpoint:
```
GET /_matrix/federation/unstable/org.matrix.msc3030/timestamp_to_event/<roomID>?ts=<timestamp>&dir=<direction>
{
    "event_id": ...
    "origin_server_ts": ...
}
```

Co-authored-by: Erik Johnston <erik@matrix.org>
2021-12-02 01:02:20 -06:00
Richard van der Hoff
f3efa0036b
Move _persist_auth_tree into FederationEventHandler (#11115)
This is just a lift-and-shift, because it fits more naturally here. We do
rename it to `process_remote_join` at the same time though.
2021-10-19 10:24:09 +01:00
Richard van der Hoff
a5d2ea3d08
Check *all* auth events for room id and rejection (#11009)
This fixes a bug where we would accept an event whose `auth_events` include
rejected events, if the rejected event was shadowed by another `auth_event`
with same `(type, state_key)`.

The approach is to pass a list of auth events into
`check_auth_rules_for_event` instead of a dict, which of course means updating
the call sites.

This is an extension of #10956.
2021-10-18 18:28:30 +01:00
Eric Eastwood
daf498e099
Fix 500 error on /messages when we accumulate more than 5 backward extremities (#11027)
Found while working on the Gitter backfill script and noticed
it only happened after we sent 7 batches, https://gitlab.com/gitterHQ/webapp/-/merge_requests/2229#note_665906390

When there are more than 5 backward extremities for a given depth,
backfill will throw an error because we sliced the extremity list
to 5 but then try to iterate over the full list. This causes
us to look for state that we never fetched and we get a `KeyError`.

Before when calling `/messages` when there are more than 5 backward extremities:
```
Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/synapse/http/server.py", line 258, in _async_render_wrapper
    callback_return = await self._async_render(request)
  File "/usr/local/lib/python3.8/site-packages/synapse/http/server.py", line 446, in _async_render
    callback_return = await raw_callback_return
  File "/usr/local/lib/python3.8/site-packages/synapse/rest/client/room.py", line 580, in on_GET
    msgs = await self.pagination_handler.get_messages(
  File "/usr/local/lib/python3.8/site-packages/synapse/handlers/pagination.py", line 396, in get_messages
    await self.hs.get_federation_handler().maybe_backfill(
  File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 133, in maybe_backfill
    return await self._maybe_backfill_inner(room_id, current_depth, limit)
  File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 386, in _maybe_backfill_inner
    likely_extremeties_domains = get_domains_from_state(states[e_id])
KeyError: '$zpFflMEBtZdgcMQWTakaVItTLMjLFdKcRWUPHbbSZJl'
```
2021-10-14 18:53:45 -05:00
Patrick Cloke
eb9ddc8c2e
Remove the deprecated BaseHandler. (#11005)
The shared ratelimit function was replaced with a dedicated
RequestRatelimiter class (accessible from the HomeServer
object).

Other properties were copied to each sub-class that inherited
from BaseHandler.
2021-10-08 07:44:43 -04:00
Patrick Cloke
d1bf5f7c9d
Strip "join_authorised_via_users_server" from join events which do not need it. (#10933)
This fixes a "Event not signed by authorising server" error when
transition room member from join -> join, e.g. when updating a
display name or avatar URL for restricted rooms.
2021-09-30 11:13:59 -04:00
Richard van der Hoff
428174f902
Split event_auth.check into two parts (#10940)
Broadly, the existing `event_auth.check` function has two parts:
 * a validation section: checks that the event isn't too big, that it has the rught signatures, etc. 
   This bit is independent of the rest of the state in the room, and so need only be done once 
   for each event.
 * an auth section: ensures that the event is allowed, given the rest of the state in the room.
   This gets done multiple times, against various sets of room state, because it forms part of
   the state res algorithm.

Currently, this is implemented with `do_sig_check` and `do_size_check` parameters, but I think
that makes everything hard to follow. Instead, we split the function in two and call each part
separately where it is needed.
2021-09-29 18:59:15 +01:00
Patrick Cloke
94b620a5ed
Use direct references for configuration variables (part 6). (#10916) 2021-09-29 06:44:15 -04:00
Richard van der Hoff
5279b9161b
Use RoomVersion objects (#10934)
Various refactors to use `RoomVersion` objects instead of room version identifiers.
2021-09-29 10:57:10 +01:00
Patrick Cloke
bb7fdd821b
Use direct references for configuration variables (part 5). (#10897) 2021-09-24 07:25:21 -04:00
Andrew Morgan
aa2c027792
Remove unnecessary parentheses around tuples returned from methods (#10889) 2021-09-23 11:59:07 +01:00