* Add `DeferredCache.get_immediate` method
A bunch of things that are currently calling `DeferredCache.get` are only
really interested in the result if it's completed. We can optimise and simplify
this case.
* Remove unused 'default' parameter to DeferredCache.get()
* another get_immediate instance
Autocommit means that we don't wrap the functions in transactions, and instead get executed directly. Introduced in #8456. This will help:
1. reduce the number of `could not serialize access due to concurrent delete` errors that we see (though there are a few functions that often cause serialization errors that we don't fix here);
2. improve the DB performance, as it no longer needs to deal with the overhead of `REPEATABLE READ` isolation levels; and
3. improve wall clock speed of these functions, as we no longer need to send `BEGIN` and `COMMIT` to the DB.
Some notes about the differences between autocommit mode and our default `REPEATABLE READ` transactions:
1. Currently `autocommit` only applies when using PostgreSQL, and is ignored when using SQLite (due to silliness with [Twisted DB classes](https://twistedmatrix.com/trac/ticket/9998)).
2. Autocommit functions may get retried on error, which means they can get applied *twice* (or more) to the DB (since they are not in a transaction the previous call would not get rolled back). This means that the functions need to be idempotent (or otherwise not care about being called multiple times). Read queries, simple deletes, and updates/upserts that replace rows (rather than generating new values from existing rows) are all idempotent.
3. Autocommit functions no longer get executed in [`REPEATABLE READ`](https://www.postgresql.org/docs/current/transaction-iso.html) isolation level, and so data can change queries, which is fine for single statement queries.
We asserted that the IDs returned by postgres sequence was greater than
any we had seen, however this is technically racey as we may update the
current positions out of order.
We now assert that the sequences are correct on startup, so the
assertion is no longer really required, so we remove them.
Currently background proccesses stream the events stream use the "minimum persisted position" (i.e. `get_current_token()`) rather than the vector clock style tokens. This is broadly fine as it doesn't matter if the background processes lag a small amount. However, in extreme cases (i.e. SyTests) where we only write to one event persister the background processes will never make progress.
This PR changes it so that the `MultiWriterIDGenerator` keeps the current position of a given instance as up to date as possible (i.e using the latest token it sees if its not in the process of persisting anything), and then periodically announces that over replication. This then allows the "minimum persisted position" to advance, albeit with a small lag.
We call `_update_stream_positions_table_txn` a lot, which is an UPSERT
that can conflict in `REPEATABLE READ` isolation level. Instead of doing
a transaction consisting of a single query we may as well run it outside
of a transaction.
Currently when using multiple event persisters we (in the worst case) don't tell clients about events until all event persisters have persisted new events after the original event. This is a suboptimal, especially if one of the event persisters goes down.
To handle this, we encode the position of each event persister in the room tokens so that we can send events to clients immediately. To reduce the size of the token we do two things:
1. We create a unique immutable persistent mapping between instance names and a generated small integer ID, which we can encode in the tokens instead of the instance name; and
2. We encode the "persisted upto position" of the room token and then only explicitly include instances that have positions strictly greater than that.
The new tokens look something like: `m3478~1.3488~2.3489`, where the first number is the min position, and the subsequent `-` separated pairs are the instance ID to positions map. (We use `.` and `~` as separators as they're URL safe and not already used by `StreamToken`).
There's no need for it to be in the dict as well as the events table. Instead,
we store it in a separate attribute in the EventInternalMetadata object, and
populate that on load.
This means that we can rely on it being correctly populated for any event which
has been persited to the database.
This is so we can tell what is going on when things are taking a while to start up.
The main change here is to ensure that transactions that are created during startup get correctly logged like normal transactions.
For negative streams we have to negate the internal stream ID before
querying the DB.
The effect of this bug was to query far too many rows, slowing start up
time, but we would correctly filter the results afterwards so there was
no ill effect.
The idea is that in future tokens will encode a mapping of instance to position. However, we don't want to include the full instance name in the string representation, so instead we'll have a mapping between instance name and an immutable integer ID in the DB that we can use instead. We'll then do the lookup when we serialize/deserialize the token (we could alternatively pass around an `Instance` type that includes both the name and ID, but that turns out to be a lot more invasive).
This was a bit unweildy for what I wanted: in particular, I wanted to assign
each measurement straight into a bucket, rather than storing an intermediate
Counter which didn't do any bucketing at all.
I've replaced it with something that is hopefully a bit easier to use.
(I'm not entirely sure what the difference between a HistogramMetricFamily and
a GaugeHistogramMetricFamily is, but given our counters can go down as well as
up the latter *sounds* more accurate?)
* Fix table scan of events on worker startup.
This happened because we assumed "new" writers had an initial stream
position of 0, so the replication code tried to fetch all events written
by the instance between 0 and the current position.
Instead, set the initial position of new writers to the current
persisted up to position, on the assumption that new writers won't have
written anything before that point.
* Consider old writers coming back as "new".
Otherwise we'd try and fetch entries between the old stale token and the
current position, even though it won't have written any rows.
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
Co-authored-by: Benjamin Koch <bbbsnowball@gmail.com>
This adds configuration flags that will match a user to pre-existing users
when logging in via OpenID Connect. This is useful when switching to
an existing SSO system.
On startup `MultiWriteIdGenerator` fetches the maximum stream ID for
each instance from the table and uses that as its initial "current
position" for each writer. This is problematic as a) it involves either
a scan of events table or an index (neither of which is ideal), and b)
if rows are being persisted out of order elsewhere while the process
restarts then using the maximum stream ID is not correct. This could
theoretically lead to race conditions where e.g. events that are
persisted out of order are not sent down sync streams.
We fix this by creating a new table that tracks the current positions of
each writer to the stream, and update it each time we finish persisting
a new entry. This is a relatively small overhead when persisting events.
However for the cache invalidation stream this is a much bigger relative
overhead, so instead we note that for invalidation we don't actually
care about reliability over restarts (as there's no caches to
invalidate) and simply don't bother reading and writing to the new table
in that particular case.
The idea is to remove some of the places we pass around `int`, where it can represent one of two things:
1. the position of an event in the stream; or
2. a token that partitions the stream, used as part of the stream tokens.
The valid operations are then:
1. did a position happen before or after a token;
2. get all events that happened before or after a token; and
3. get all events between two tokens.
(Note that we don't want to allow other operations as we want to change the tokens to be vector clocks rather than simple ints)
When updating room_stats_state, we try to check for null bytes slipping
in to the
content for state events. It turns out we had added guest_access as a
field to
room_stats_state without including it in the null byte check.
Lo and behold, a null byte in a m.room.guest_access event then breaks
room_stats_state
updates.
This PR adds the check for guest_access. A further PR will improve this
function so that this hopefully does not happen again in future.
Fixes: #8359
Trying to reactivate a user with the admin API (`PUT /_synapse/admin/v2/users/<user_name>`) causes an internal server error.
Seems to be a regression in #8033.
==============================
In addition to the below, Synapse 1.20.0rc5 also includes the bug fix that was included in 1.19.3.
Features
--------
- Add flags to the `/versions` endpoint for whether new rooms default to using E2EE. ([\#8343](https://github.com/matrix-org/synapse/issues/8343))
Bugfixes
--------
- Fix rate limiting of federation `/send` requests. ([\#8342](https://github.com/matrix-org/synapse/issues/8342))
- Fix a longstanding bug where back pagination over federation could get stuck if it failed to handle a received event. ([\#8349](https://github.com/matrix-org/synapse/issues/8349))
Internal Changes
----------------
- Blacklist [MSC2753](https://github.com/matrix-org/matrix-doc/pull/2753) SyTests until it is implemented. ([\#8285](https://github.com/matrix-org/synapse/issues/8285))
-----BEGIN PGP SIGNATURE-----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=tT5/
-----END PGP SIGNATURE-----
Merge tag 'v1.20.0rc5' into develop
Synapse 1.20.0rc5 (2020-09-18)
==============================
In addition to the below, Synapse 1.20.0rc5 also includes the bug fix that was included in 1.19.3.
Features
--------
- Add flags to the `/versions` endpoint for whether new rooms default to using E2EE. ([\#8343](https://github.com/matrix-org/synapse/issues/8343))
Bugfixes
--------
- Fix rate limiting of federation `/send` requests. ([\#8342](https://github.com/matrix-org/synapse/issues/8342))
- Fix a longstanding bug where back pagination over federation could get stuck if it failed to handle a received event. ([\#8349](https://github.com/matrix-org/synapse/issues/8349))
Internal Changes
----------------
- Blacklist [MSC2753](https://github.com/matrix-org/matrix-doc/pull/2753) SyTests until it is implemented. ([\#8285](https://github.com/matrix-org/synapse/issues/8285))
Signed-off-by: Olivier Wilkinson (reivilibre) <olivier@librepush.net>
Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
* Fix _set_destination_retry_timings
This came about because the code assumed that retry_interval
could not be NULL — which has been challenged by catch-up.
Instead of just using the most recent extremities let's pick the
ones that will give us results that the pagination request cares about,
i.e. pick extremities only if they have a smaller depth than the
pagination token.
This is useful when we fail to backfill an extremity, as we no longer
get stuck requesting that same extremity repeatedly.
slots use less memory (and attribute access is faster) while slightly
limiting the flexibility of the class attributes. This focuses on objects
which are instantiated "often" and for short periods of time.
This is *not* ready for production yet. Caveats:
1. We should write some tests...
2. The stream token that we use for events can get stalled at the minimum position of all writers. This means that new events may not be processed and e.g. sent down sync streams if a writer isn't writing or is slow.
This fixes an issue where different methods (crop/scale) overwrite each other.
This first tries the new path. If that fails and we are looking for a
remote thumbnail, it tries the old path. If that still isn't found, it
continues as normal.
This should probably be removed in the future, after some of the newer
thumbnails were generated with the new path on most deployments. Then
the overhead should be minimal if the other thumbnails need to be
regenerated.
Signed-off-by: Nicolas Werner <nicolas.werner@hotmail.de>
The intention here is to change `StreamToken.room_key` to be a `RoomStreamToken` in a future PR, but that is a big enough change without this refactoring too.
Fixes https://github.com/matrix-org/synapse/issues/8238
Alongside the delta file, some changes were also necessary to the codebase to remove references to the now defunct `populate_stats_process_rooms_2` background job. Thankfully the latter doesn't seem to have made it into any documentation yet :)
* Fixup `ALTER TABLE` database queries
Make the new columns nullable, because doing otherwise can wedge a
server with a big database, as setting a default value rewrites the
table.
* Switch back to using the notifications count in the push badge
Clients are likely to be confused if we send a push but the badge count
is the unread messages one, and not the notifications one.
* Changelog
This is *not* ready for production yet. Caveats:
1. We should write some tests...
2. The stream token that we use for events can get stalled at the minimum position of all writers. This means that new events may not be processed and e.g. sent down sync streams if a writer isn't writing or is slow.
* Add shared_rooms api
* Add changelog
* Add .
* Wrap response in {"rooms": }
* linting
* Add unstable_features key
* Remove options from isort that aren't part of 5.x
`-y` and `-rc` are now default behaviour and no longer exist.
`dont-skip` is no longer required
https://timothycrosley.github.io/isort/CHANGELOG/#500-penny-july-4-2020
* Update imports to make isort happy
* Add changelog
* Update tox.ini file with correct invocation
* fix linting again for isort
* Vendor prefix unstable API
* Fix to match spec
* import Codes
* import Codes
* Use FORBIDDEN
* Update changelog.d/7785.feature
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
* Implement get_shared_rooms_for_users
* a comma
* trailing whitespace
* Handle the easy feedback
* Switch to using runInteraction
* Add tests
* Feedback
* Seperate unstable endpoint from v2
* Add upgrade node
* a line
* Fix style by adding a blank line at EOF.
* Update synapse/storage/databases/main/user_directory.py
Co-authored-by: Tulir Asokan <tulir@maunium.net>
* Update synapse/storage/databases/main/user_directory.py
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
* Update UPGRADE.rst
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
* Fix UPGRADE/CHANGELOG unstable paths
unstable unstable unstable
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
Co-authored-by: Tulir Asokan <tulir@maunium.net>
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
Co-authored-by: Tulir Asokan <tulir@maunium.net>
* Move `get_devices_with_keys_by_user` to `EndToEndKeyWorkerStore`
this seems a better fit for it.
This commit simply moves the existing code: no other changes at all.
* Rename `get_devices_with_keys_by_user`
to better reflect what it does.
* get_device_stream_token abstract method
To avoid referencing fields which are declared in the derived classes, make
`get_device_stream_token` abstract, and define that in the classes which define
`_device_list_id_gen`.
... and to show that it does something slightly different to
`_get_e2e_device_keys_txn`.
`include_all_devices` and `include_deleted_devices` were never used (and
`include_deleted_devices` was broken, since that would cause `None`s in the
result which were not handled in the loop below.
Add some typing too.
* Don't raise session_id errors on submit_token if request_token_inhibit_3pid_errors is set
* Changelog
* Also wait some time before responding to /requestToken
* Incorporate review
* Update synapse/storage/databases/main/registration.py
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
* Incorporate review
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
It's just a thin wrapper around two ID gens to make `get_current_token`
and `get_next` return tuples. This can easily be replaced by calling the
appropriate methods on the underlying ID gens directly.
The function is used for two purposes: 1) for subscribers of streams to
get a token they can use to get further updates with, and 2) for
replication to track position of the writers of the stream.
For streams with a single writer the two scenarios produce the same
result, however the situation becomes complicated for streams with
multiple writers. The current `MultiWriterIdGenerator` does not
correctly handle the first case (which is not an issue as its only used
for the `caches` stream which nothing subscribes to outside of
replication).
Duplicating function signatures between server.py and server.pyi is
silly. This commit changes that by changing all `build_*` methods to
`get_*` methods and changing the `_make_dependency_method` to work work
as a descriptor that caches the produced value.
There are some changes in other files that were made to fix the typing
in server.py.
`StatsHandler` handles updates to the `current_state_delta_stream`, and updates room stats such as the amount of state events, joined users, etc.
However, it counts every new join membership as a new user entering a room (and that user being in another room), whereas it's possible for a user's membership status to go from join -> join, for instance when they change their per-room profile information.
This PR adds a check for join->join membership transitions, and bails out early, as none of the further checks are necessary at that point.
Due to this bug, membership stats in many rooms have ended up being wildly larger than their true values. I am not sure if we also want to include a migration step which recalculates these statistics (possibly using the `_populate_stats_process_rooms` bg update).
Bug introduced in the initial implementation https://github.com/matrix-org/synapse/pull/4338.
It serves no purpose and updating everytime we write to the device inbox
stream means all such transactions will conflict, causing lots of
transaction failures and retries.
When considering rooms to clean up in `delete_old_current_state_events`, skip
rooms which we are creating, which otherwise look a bit like rooms we have
left.
Fixes#7834.
As far as I can tell from the sentry logs, the only time this has actually done
anything in the last two years is when we had two master workers running at
once, and even then, it made a bit of a mess of it (see
https://github.com/matrix-org/synapse/issues/7845#issuecomment-658238739).
Generally I feel like this code is doing more harm than good.
Fixes#2181.
The basic premise is that, when we
fail to reject an invite via the remote server, we can generate our own
out-of-band leave event and persist it as an outlier, so that we have something
to send to the client.
This table is no longer used, so we may as well stop populating it. Removing it
would prevent people rolling back to older releases of Synapse, so that can
happen in a future release.
The CI appears to use the latest version of isort, which is a problem when isort gets a major version bump. Rather than try to pin the version, I've done the necessary to make isort5 happy with synapse.
* Always return an unread_count in get_unread_event_push_actions_by_room_for_user
* Don't always expect unread_count to be there so we don't take out sync entirely if something goes wrong
The aim here is to make it easier to reason about when streams are limited and when they're not, by moving the logic into the database functions themselves. This should mean we can kill of `db_query_to_update_function` function.
* Ensure account data stream IDs are unique.
The account data stream is shared between three tables, and the maximum
allocated ID was tracked in a dedicated table. Updating the max ID
happened outside the transaction that allocated the ID, leading to a
race where if the server was restarted then the same ID could be
allocated but the max ID failed to be updated, leading it to be reused.
The ID generators have support for tracking across multiple tables, so
we may as well use that instead of a dedicated table.
* Fix bug in account data replication stream.
If the same stream ID was used in both global and room account data then
the getting updates for the replication stream would fail due to
`heapq.merge(..)` trying to compare a `str` with a `None`. (This is
because you'd have two rows like `(534, '!room')` and `(534, None)` from
the room and global account data tables).
Fix is just to order by stream ID, since we don't rely on the ordering
beyond that. The bug where stream IDs can be reused should be fixed now,
so this case shouldn't happen going forward.
Fixes#7617
The query keeps showing up in my slow query log.
This changes the plan under the top-level Sort node from
```
WindowAgg (cost=280335.88..292963.15 rows=561212 width=80) (actual time=138.651..160.562 rows=27112 loops=1)
-> Sort (cost=280335.88..281738.91 rows=561212 width=84) (actual time=138.597..140.622 rows=27112 loops=1)
Sort Key: state_groups_state.type, state_groups_state.state_key, state_groups_state.state_group
Sort Method: quicksort Memory: 4581kB
-> Nested Loop (cost=2.83..226745.22 rows=561212 width=84) (actual time=21.548..47.657 rows=27112 loops=1)
-> HashAggregate (cost=2.27..3.28 rows=101 width=8) (actual time=21.526..21.535 rows=20 loops=1)
Group Key: state.state_group
-> CTE Scan on state (cost=0.00..2.02 rows=101 width=8) (actual time=21.280..21.493 rows=20 loops=1)
-> Index Scan using state_groups_state_type_idx on state_groups_state (cost=0.56..2189.40 rows=5557 width=84) (actual time=0.005..0.991 rows=1356 loops=20)
Index Cond: (state_group = state.state_group)
```
to
```
Nested Loop (cost=2.83..226745.22 rows=561212 width=84) (actual time=24.194..52.834 rows=27112 loops=1)
-> HashAggregate (cost=2.27..3.28 rows=101 width=8) (actual time=24.130..24.138 rows=20 loops=1)
Group Key: state.state_group
-> CTE Scan on state (cost=0.00..2.02 rows=101 width=8) (actual time=23.887..24.113 rows=20 loops=1)
-> Index Scan using state_groups_state_type_idx on state_groups_state (cost=0.56..2189.40 rows=5557 width=84) (actual time=0.016..1.159 rows=1356 loops=20)
Index Cond: (state_group = state.state_group)
```
This cuts the execution time from ~190ms to ~130ms, i.e. a reduction
of ~30%.
The full plans are visualised at https://explain.depesz.com/s/WpbT and
https://explain.depesz.com/s/KlEk
Signed-off-by: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>
The bg update never managed to complete, because it kept being interrupted by
transactions which want to take a lock.
Just doing it in the foreground isn't that bad, and is a good deal simpler.
we can use `make_in_list_sql_clause` rather than doing our own half-baked
equivalent, which has the benefit of working just fine with empty lists.
(This has quite a lot of tests, so I think it's pretty safe)
The idea here is that if an instance persists an event via the replication HTTP API it can return before we receive that event over replication, which can lead to races where code assumes that persisting an event immediately updates various caches (e.g. current state of the room).
Most of Synapse doesn't hit such races, so we don't do the waiting automagically, instead we do so where necessary to avoid unnecessary delays. We may decide to change our minds here if it turns out there are a lot of subtle races going on.
People probably want to look at this commit by commit.
When a call to `user_device_resync` fails, we don't currently mark the remote user's device list as out of sync, nor do we retry to sync it.
https://github.com/matrix-org/synapse/pull/6776 introduced some code infrastructure to mark device lists as stale/out of sync.
This commit uses that code infrastructure to mark device lists as out of sync if processing an incoming device list update makes the device handler realise that the device list is out of sync, but we can't resync right now.
It also adds a looping call to retry all failed resync every 30s. This shouldn't cause too much spam in the logs as this commit also removes the "Failed to handle device list update for..." warning logs when catching `NotRetryingDestination`.
Fixes#7418
`_is_server_still_joined` will throw if it is given state updates with non-user ID state keys with local user leaves. This is actually rarely a problem since local leaves almost always get persisted by themselves.
(I discovered this on a branch that was otherwise broken, so I haven't seen this in the wild)
Make sure that the AccountDataStream presents complete updates, in the right
order.
This is much the same fix as #7337 and #7358, but applied to a different stream.
This is required as both event persistence and the background update needs access to this function. It should be perfectly safe for two workers to write to that table at the same time.
