eb29a31550
Should fix the following issues or make a lot less worse when using Postgres: The main issue behind #2911: The client gives up after a certain time, causing a cascade of context errors, because the response couldn't be built up fast enough. This mostly happens on accounts with many rooms, due to the inefficient way we're getting recent events and current state For #2777: The queries for getting the membership events for history visibility were being executed for each room (I think 185?), resulting in a whooping 2k queries for membership events. (Getting the statesnapshot -> block nids -> actual wanted membership event) Both should now be better by: - Using a LATERAL join to get all recent events for all joined rooms in one go (TODO: maybe do the same for room summary and current state etc) - If we're lazy loading on initial syncs, we're now not getting the whole current state, just to drop the majority of it because we're lazy loading members - we add a filter to exclude membership events on the first call to `CurrentState`. - Using an optimized query to get the membership events needed to calculate history visibility --------- Co-authored-by: kegsay <kegan@matrix.org> |
||
---|---|---|
.. | ||
acls | ||
api | ||
auth | ||
internal | ||
inthttp | ||
producers | ||
state | ||
storage | ||
types | ||
version | ||
README.md | ||
roomserver_test.go | ||
roomserver.go |
RoomServer
RoomServer Internals
Numeric IDs
To save space matrix string identifiers are mapped to local numeric IDs. The numeric IDs are more efficient to manipulate and use less space to store. The numeric IDs are never exposed in the API the room server exposes. The numeric IDs are converted to string IDs before they leave the room server. The numeric ID for a string ID is never 0 to avoid being confused with go's default zero value. Zero is used to indicate that there was no corresponding string ID. Well-known event types and event state keys are preassigned numeric IDs.
State Snapshot Storage
The room server stores the state of the matrix room at each event. For efficiency the state is stored as blocks of 3-tuples of numeric IDs for the event type, event state key and event ID. For further efficiency the state snapshots are stored as the combination of up to 64 these blocks. This allows blocks of the room state to be reused in multiple snapshots.
The resulting database tables look something like this:
+-------------------------------------------------------------------+
| Events |
+---------+-------------------+------------------+------------------+
| EventNID| EventTypeNID | EventStateKeyNID | StateSnapshotNID |
+---------+-------------------+------------------+------------------+
| 1 | m.room.create 1 | "" 1 | <nil> 0 |
| 2 | m.room.member 2 | "@user:foo" 2 | <nil> 0 |
| 3 | m.room.member 2 | "@user:bar" 3 | {1,2} 1 |
| 4 | m.room.message 3 | <nil> 0 | {1,2,3} 2 |
| 5 | m.room.member 2 | "@user:foo" 2 | {1,2,3} 2 |
| 6 | m.room.message 3 | <nil> 0 | {1,3,6} 3 |
+---------+-------------------+------------------+------------------+
+----------------------------------------+
| State Snapshots |
+-----------------------+----------------+
| EventStateSnapshotNID | StateBlockNIDs |
+-----------------------+----------------|
| 1 | {1} |
| 2 | {1,2} |
| 3 | {1,2,3} |
+-----------------------+----------------+
+-----------------------------------------------------------------+
| State Blocks |
+---------------+-------------------+------------------+----------+
| StateBlockNID | EventTypeNID | EventStateKeyNID | EventNID |
+---------------+-------------------+------------------+----------+
| 1 | m.room.create 1 | "" 1 | 1 |
| 1 | m.room.member 2 | "@user:foo" 2 | 2 |
| 2 | m.room.member 2 | "@user:bar" 3 | 3 |
| 3 | m.room.member 2 | "@user:foo" 2 | 6 |
+---------------+-------------------+------------------+----------+