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.
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.
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>
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.
... 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>
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).
Calls `self.get_success` on all deferred methods instead of abusing `self.pump()`. This has the benefit of working with coroutines, as well as checking that method execution completed successfully.
There are also a few small cleanups that I made in the process.
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.
==============================
Bugfixes
--------
- Fix a long-standing bug which could cause messages not to be sent over federation, when state events with state keys matching user IDs (such as custom user statuses) were received. ([\#7376](https://github.com/matrix-org/synapse/issues/7376))
- Restore compatibility with non-compliant clients during the user interactive authentication process, fixing a problem introduced in v1.13.0rc1. ([\#7483](https://github.com/matrix-org/synapse/issues/7483))
Internal Changes
----------------
- Fix linting errors in new version of Flake8. ([\#7470](https://github.com/matrix-org/synapse/issues/7470))
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEEv27Axt/F4vrTL/8QOSor00I9eP8FAl69IQ8ACgkQOSor00I9
eP87lAf8DK+v6cs2U0BoD5opzQ7ZazJT6JYTmnMBaTzHU6Wx20V2ttkF7Vpwm3WU
Zsz0048tdYtHFyYBQ1kF5RNIBBJwV8SA/QUcPkR7FVpwZMLR2q4aJn0EE7kC9OMf
tYsmdbHeBdyfLXpXzazxWlgHquLyEIt52ykAcCphjx/Jl2fAExFEhtfsxpECoJ2f
8Dqhjg3WFjd6QWU6AFkElbwHUYCdIWdJOcsC8N1p8OvBmDz5QXv/RlYipHE00Cpx
QQQOgEjdRc6dlz2mbetMklnfII3p2kO9bzNdmEpOzT0Zt7nFaGdntW4I1QA0yJfa
gows9bYMzhqYk7YSiyTYOZ4qyavVtw==
=N/zZ
-----END PGP SIGNATURE-----
Merge tag 'v1.13.0rc2' into develop
Synapse 1.13.0rc2 (2020-05-14)
==============================
Bugfixes
--------
- Fix a long-standing bug which could cause messages not to be sent over federation, when state events with state keys matching user IDs (such as custom user statuses) were received. ([\#7376](https://github.com/matrix-org/synapse/issues/7376))
- Restore compatibility with non-compliant clients during the user interactive authentication process, fixing a problem introduced in v1.13.0rc1. ([\#7483](https://github.com/matrix-org/synapse/issues/7483))
Internal Changes
----------------
- Fix linting errors in new version of Flake8. ([\#7470](https://github.com/matrix-org/synapse/issues/7470))
Fix a bug where the `get_joined_users` cache could be corrupted by custom
status events (or other state events with a state_key matching the user ID).
The bug was introduced by #2229, but has largely gone unnoticed since then.
Fixes#7099, #7373.
The aim here is to get to a stage where we have a `PersistEventStore` that holds all the write methods used during event persistence, so that we can take that class out of the `DataStore` mixin and instansiate it separately. This will allow us to instansiate it on processes other than master, while also ensuring it is only available on processes that are configured to write to events stream.
This is a bit of an architectural change, where we end up with multiple classes per data store (rather than one per data store we have now). We end up having:
1. Storage classes that provide high level APIs that can talk to multiple data stores.
2. Data store modules that consist of classes that must point at the same database instance.
3. Classes in a data store that can be instantiated on processes depending on config.
* Add 'device_lists_outbound_pokes' as extra table.
This makes sure we check all the relevant tables to get the current max
stream ID.
Currently not doing so isn't problematic as the max stream ID in
`device_lists_outbound_pokes` is the same as in `device_lists_stream`,
however that will change.
* Change device lists stream to have one row per id.
This will make it possible to process the streams more incrementally,
avoiding having to process large chunks at once.
* Change device list replication to match new semantics.
Instead of sending down batches of user ID/host tuples, send down a row
per entity (user ID or host).
* Newsfile
* Remove handling of multiple rows per ID
* Fix worker handling
* Comments from review
It was originally implemented by pulling the full auth chain of all
state sets out of the database and doing set comparison. However, that
can take a lot work if the state and auth chains are large.
Instead, lets try and fetch the auth chains at the same time and
calculate the difference on the fly, allowing us to bail early if all
the auth chains converge. Assuming that the auth chains do converge more
often than not, this should improve performance. Hopefully.
There are quite a few places that we assume that a redaction event has a
corresponding `redacts` key, which is not always the case. So lets
cheekily make it so that event.redacts just returns None instead.
Currently we rely on `current_state_events` to figure out what rooms a
user was in and their last membership event in there. However, if the
server leaves the room then the table may be cleaned up and that
information is lost. So lets add a table that separately holds that
information.
This makes it easier to use in an async/await world.
Also fixes a bug where cache descriptors would occaisonally return a raw
value rather than a deferred.
We have set the max retry interval to a value larger than a postgres or
sqlite int can hold, which caused exceptions when updating the
destinations table.
To fix postgres we need to change the column to a bigint, and for sqlite
we lower the max interval to 2**62 (which is still incredibly long).
This is a) simpler than querying user_ips directly and b) means we can
purge older entries from user_ips without losing the required info.
The storage functions now no longer return the access_token, since it
was unused.
Adds new config option `cleanup_extremities_with_dummy_events` which
periodically sends dummy events to rooms with more than 10 extremities.
THIS IS REALLY EXPERIMENTAL.
