This is the final step for supporting multiple OIDC providers concurrently.
First of all, we reorganise the config so that you can specify a list of OIDC providers, instead of a single one. Before:
oidc_config:
enabled: true
issuer: "https://oidc_provider"
# etc
After:
oidc_providers:
- idp_id: prov1
issuer: "https://oidc_provider"
- idp_id: prov2
issuer: "https://another_oidc_provider"
The old format is still grandfathered in.
With that done, it's then simply a matter of having OidcHandler instantiate a new OidcProvider for each configured provider.
Protecting media stops it from being quarantined when
e.g. all media in a room is quarantined. This is useful
for sticker packs and other media that is uploaded by
server administrators, but used by many people.
`distutils` is pretty much deprecated these days, and replaced with
`setuptools`. It's also annoying because it's you can't `pip install` it, and
it's hard to figure out which debian package we should depend on to make sure
it's there.
Since we only use it for a tiny function anyway, let's just vendor said
function into our codebase.
* make the OIDC bits of the test work at a higher level - via the REST api instead of poking the OIDCHandler directly.
* Move it to test_login.py, where I think it fits better.
Again in preparation for handling more than one OIDC provider, add a new caveat to the macaroon used as an OIDC session cookie, which remembers which OIDC provider we are talking to. In future, when we get a callback, we'll need it to make sure we talk to the right IdP.
As part of this, I'm adding an idp_id and idp_name field to the OIDC configuration object. They aren't yet documented, and we'll just use the old values by default.
The idea here is that we will have an instance of OidcProvider for each
configured IdP, with OidcHandler just doing the marshalling of them.
For now it's still hardcoded with a single provider.
A reactor was being passed instead of a whitelist for the BlacklistingAgentWrapper
used by the WellyKnownResolver. This coulld cause exceptions when attempting to
connect to IP addresses that are blacklisted, but in reality this did not have any
observable affect since this code is not used for IP literals.
If a user tries to do UI Auth via SSO, but uses the wrong account on the SSO
IdP, try to give them a better error.
Previously, the UIA would claim to be successful, but then the operation in
question would simply fail with "auth fail". Instead, serve up an error page
which explains the failure.
This checks that the domain given to `DomainSpecificString.is_valid` (e.g.
`UserID`, `RoomAlias`, etc.) is of a valid form. Previously some validation
was done on the localpart (e.g. the sigil), but not the domain portion.
Some light refactoring of OidcHandler, in preparation for bigger things:
* remove inheritance from deprecated BaseHandler
* add an object to hold the things that go into a session cookie
* factor out a separate class for manipulating said cookies
If we have integrations with multiple identity providers, when the user does a UI Auth, we need to redirect them to the right one.
There are a few steps to this. First of all we actually need to store the userid of the user we are trying to validate in the UIA session, since the /auth/sso/fallback/web request is unauthenticated.
Then, once we get the /auth/sso/fallback/web request, we can fish the user id out of the session, and use it to look up the external id mappings, and hence pick an SSO provider for them.
SynapseRequest is in danger of becoming a bit of a dumping-ground for "useful stuff relating to Requests",
which isn't really its intention (its purpose is to override render, finished and connectionLost to set up the
LoggingContext and write the right entries to the request log).
Putting utility functions inside SynapseRequest means that lots of our code ends up requiring a
SynapseRequest when there is nothing synapse-specific about the Request at all, and any old
twisted.web.iweb.IRequest will do. This increases code coupling and makes testing more difficult.
In short: move get_user_agent out to a utility function.
You can't continue using a transaction once an exception has been
raised, so catching and dropping the error here is pointless and just
causes more errors.
I'm not even sure what this was supposed to do, but the fact it has python2isms
and nobody has noticed suggests it's not terribly important.
It doesn't seem to have been used since ff23e5ba37.
* Implement CasHandler.handle_redirect_request
... to make it match OidcHandler and SamlHandler
* Clean up interface for OidcHandler.handle_redirect_request
Make it accept `client_redirect_url=None`.
* Clean up interface for `SamlHandler.handle_redirect_request`
... bring it into line with CAS and OIDC by making it take a Request parameter,
move the magic for `client_redirect_url` for UIA into the handler, and fix the
return type to be a `str` rather than a `bytes`.
* Define a common protocol for SSO auth provider impls
* Give SsoIdentityProvider an ID and register them
* Combine the SSO Redirect servlets
Now that the SsoHandler knows about the identity providers, we can combine the
various *RedirectServlets into a single implementation which delegates to the
right IdP.
* changelog
The `RoomDirectoryFederationTests` tests were not being run unless explicitly called as an `__init__.py` file was not present in `tests/federation/transport/`. Thus the folder was not a python module, and `trial` did not look inside for any test cases to run. This was found while working on #6739.
This PR adds a `__init__.py` and also fixes the test in a couple ways:
- Switch to subclassing `unittest.FederatingHomeserverTestCase` instead, which sets up federation endpoints for us.
- Supply a `federation_auth_origin` to `make_request` in order to more act like the request is coming from another server, instead of just an unauthenicated client requesting a federation endpoint.
I found that the second point makes no difference to the test passing, but felt like the right thing to do if we're testing over federation.
This adds an admin API that allows a server admin to get power in a room if a local user has power in a room. Will also invite the user if they're not in the room and its a private room. Can specify another user (rather than the admin user) to be granted power.
Co-authored-by: Matthew Hodgson <matthew@matrix.org>
This had two effects 1) it'd give the wrong answer and b) would iterate
*all* power levels in the auth chain of each event. The latter of which
can be *very* expensive for certain types of IRC bridge rooms that have
large numbers of power level changes.
The final part (for now) of my work to implement a username picker in synapse itself. The idea is that we allow
`UsernameMappingProvider`s to return `localpart=None`, in which case, rather than redirecting the browser
back to the client, we redirect to a username-picker resource, which allows the user to enter a username.
We *then* complete the SSO flow (including doing the client permission checks).
The static resources for the username picker itself (in
https://github.com/matrix-org/synapse/tree/rav/username_picker/synapse/res/username_picker)
are essentially lifted wholesale from
https://github.com/matrix-org/matrix-synapse-saml-mozilla/tree/master/matrix_synapse_saml_mozilla/res.
As the comment says, we might want to think about making them customisable, but that can be a follow-up.
Fixes#8876.
Fixes a bug that deactivated users appear in the directory when their profile information was updated.
To change profile information of deactivated users is neccesary for example you will remove displayname or avatar.
But they should not appear in directory. They are deactivated.
Co-authored-by: Erik Johnston <erikj@jki.re>
This is another part of my work towards fixing #8876. It moves some of the logic currently in the SAML and OIDC handlers - in particular the call to `AuthHandler.complete_sso_login` down into the `SsoHandler`.
* move simple_async_mock to test_utils
... so that it can be re-used
* Remove references to `SamlHandler._map_saml_response_to_user` from tests
This method is going away, so we can no longer use it as a test point. Instead,
factor out a higher-level method which takes a SAML object, and verify correct
behaviour by mocking out `AuthHandler.complete_sso_login`.
* changelog
* Remove references to handler._auth_handler
(and replace them with hs.get_auth_handler)
* Factor out a utility function for building Requests
* Remove mocks of `OidcHandler._map_userinfo_to_user`
This method is going away, so mocking it out is no longer a valid approach.
Instead, we mock out lower-level methods (eg _remote_id_from_userinfo), or
simply allow the regular implementation to proceed and update the expectations
accordingly.
* Remove references to `OidcHandler._map_userinfo_to_user` from tests
This method is going away, so we can no longer use it as a test point. Instead
we build mock "callback" requests which we pass into `handle_oidc_callback`,
and verify correct behaviour by mocking out `AuthHandler.complete_sso_login`.
* Factor out _call_attribute_mapper and _register_mapped_user
This is mostly an attempt to simplify `get_mxid_from_sso`.
* Move mapping_lock down into SsoHandler.
It seems that letting CircleCI use its default docker version (17.09.0-ce,
apparently) did not interact well with multiarch builds: in particular, we saw
weird effects where running an amd64 build at the same time as an arm64 build
caused the arm64 builds to fail with:
Error while loading /usr/sbin/dpkg-deb: No such file or directory
The idea is that the parse_config method of extension modules can raise either a ConfigError or a JsonValidationError,
and it will be magically turned into a legible error message. There's a few components to it:
* Separating the "path" and the "message" parts of a ConfigError, so that we can fiddle with the path bit to turn it
into an absolute path.
* Generally improving the way ConfigErrors get printed.
* Passing in the config path to load_module so that it can wrap any exceptions that get caught appropriately.
* SsoHandler: remove inheritance from BaseHandler
* Simplify the flow for SSO UIA
We don't need to do all the magic for mapping users when we are doing UIA, so
let's factor that out.
* Call set_avatar_url with target_user, not user_id
Fixes https://github.com/matrix-org/synapse/issues/8871
* Create 8872.bugfix
* Update synapse/rest/admin/users.py
Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
* Testing
* Update changelog.d/8872.bugfix
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>
This removes the version pin of the `prometheus_client` dependency, in direct response to #8831. If merged, this will close#8831
As far as I can tell, no other changes are needed, but as I'm no synapse expert, I'm relying heavily on CI and maintainer reviews for this. My very primitive test of synapse with prometheus_client v0.9.0 on my home server didn't bring up any issues, so we'll see what happens.
Signed-off-by: Jordan Bancino
Replaces the `federation_ip_range_blacklist` configuration setting with an
`ip_range_blacklist` setting with wider scope. It now applies to:
* Federation
* Identity servers
* Push notifications
* Checking key validitity for third-party invite events
The old `federation_ip_range_blacklist` setting is still honored if present, but
with reduced scope (it only applies to federation and identity servers).
We do state res with unpersisted events when calculating the new current state of the room, so that should be the only thing impacted. I don't think this is tooooo big of a deal as:
1. the next time a state event happens in the room the current state should correct itself;
2. in the common case all the unpersisted events' auth events will be pulled in by other state, so will still return the correct result (or one which is sufficiently close to not affect the result); and
3. we mostly use the state at an event to do important operations, which isn't affected by this.
The official dashboard uses data from these rules, but they were never added to the synapse-v2.rules. They are mentioned in this issue: https://github.com/matrix-org/synapse/issues/7917#issuecomment-661330409, but never got added to the rules.
Adding them results in all graphs in the "Event persist rate" section to function as intended.
Signed-off-by: Johanna Dorothea Reichmann <transcaffeine@finallycoffee.eu>
This was broken in #8801 when abstracting code shared with OIDC.
After this change both SAML and OIDC have a concept of
grandfathering users, but with different implementations.
This PR adds a `room_version` argument to the `RestHelper`'s `create_room_as` function for tests. I plan to use this for testing knocking, which currently uses an unstable room version.
The spec requires synapse to support `identifier` dicts for `m.login.password`
user-interactive auth, which it did not (instead, it required an undocumented
`user` parameter.)
To fix this properly, we need to pull the code that interprets `identifier`
into `AuthHandler.validate_login` so that it can be called from the UIA code.
Fixes#5665.
It's important that we make sure our background updates happen in a defined
order, to avoid disasters like #6923.
Add an ordering to all of the background updates that have landed since #7190.
This applies even if the feature is disabled at the server level with `allow_per_room_profiles`.
The server notice not being a real user it doesn't have an user profile.
This PR adds a new config option to the `push` section of the homeserver config, `group_unread_count_by_room`. By default Synapse will group push notifications by room (so if you have 1000 unread messages, if they lie in 55 rooms, you'll see an unread count on your phone of 55).
However, it is also useful to be able to send out the true count of unread messages if desired. If `group_unread_count_by_room` is set to `false`, then with the above example, one would see an unread count of 1000 (email anyone?).
This PR grew out of #6739, and adds typing to some method arguments
You'll notice that there are a lot of `# type: ignores` in here. This is due to the base methods not matching the overloads here. This is necessary to stop mypy complaining, but a better solution is #8828.
We can get a SIGHUP at any point, including times where we are not in a
sane state. By deferring calling the handlers until the next reactor
tick we ensure that we don't get unexpected conflicts, e.g. trying to
flush logs from the signal handler while the code was in the process of
writing a log entry.
Fixes#8769.
When server URL provided to register_new_matrix_user includes path
component (e.g. "http://localhost:8008/"), the command fails with
"ERROR! Received 400 Bad Request". Stripping trailing slash from the
server_url command argument makes sure combined endpoint URL remains
valid.
Signed-off-by: Dmitry Borodaenko angdraug@debian.org
This is another PR that grew out of #6739.
The existing code for checking whether a user is currently invited to a room when they want to leave the room looks like the following:
f737368a26/synapse/handlers/room_member.py (L518-L540)
It calls `get_invite_for_local_user_in_room`, which will actually query *all* rooms the user has been invited to, before iterating over them and matching via the room ID. It will then return a tuple of a lot of information which we pull the event ID out of.
I need to do a similar check for knocking, but this code wasn't very efficient. I then tried to write a different implementation using `StateHandler.get_current_state` but this actually didn't work as we haven't *joined* the room yet - we've only been invited to it. That means that only certain tables in Synapse have our desired `invite` membership state. One of those tables is `local_current_membership`.
So I wrote a store method that just queries that table instead
Synctl did not check if a worker thread was already running when using
`synctl start` and would naively start a fresh copy. This would
sometimes lead to cases where many duplicate copies of a single worker
would run.
This fix adds a pid check when starting worker threads and synctl will
now refuse to start individual workers if they're already running.
When using `add_header` nginx will literally add a header. If a
`content-type` header is already configured (for example through a
server wide default), this means we end up with 2 content-type headers,
like so:
```
content-type: text/html
content-type: application/json
access-control-allow-origin: *
```
That doesn't make sense. Instead, we want the content type of that
block to only be `application/json` which we can achieve using
`default_type` instead.
Signed-off-by: Daniele Sluijters <daenney@users.noreply.github.com>
Checks that the localpart returned by mapping providers for SAML and
OIDC are valid before registering new users.
Extends the OIDC tests for existing users and invalid data.
* Consistently use room_id from federation request body
Some federation APIs have a redundant `room_id` path param (see
https://github.com/matrix-org/matrix-doc/issues/2330). We should make sure we
consistently use either the path param or the body param, and the body param is
easier.
* Kill off some references to "context"
Once upon a time, "rooms" were known as "contexts". I think this kills of the
last references to "contexts".
* Make this line debug (it's noisy)
* Don't include from_key for presence if we are at 0
* Limit read receipts for all rooms to 100
* changelog.d/8744.bugfix
* Allow from_key to be None
* Update 8744.bugfix
* The from_key is superflous
* Update comment
`_locally_reject_invite` generates an out-of-band membership event which can be passed to clients, but not other homeservers.
This is used when we fail to reject an invite over federation. If this happens, we instead just generate a leave event locally and send it down /sync, allowing clients to reject invites even if we can't reach the remote homeserver.
A similar flow needs to be put in place for rescinding knocks. If we're unable to contact any remote server from the room we've tried to knock on, we'd still like to generate and store the leave event locally. Hence the need to reuse, and thus generalise, this method.
Separated from #6739.
The root resource isn't necessarily a JsonResource, so rename this method
accordingly, and update a couple of test classes to use the method rather than
directly manipulating self.resource.
There's a handy function called maybe_store_room_on_invite which allows us to create an entry in the rooms table for a room and its version for which we aren't joined to yet, but we can reference when ingesting events about.
This is currently used for invites where we receive some stripped state about the room and pass it down via /sync to the client, without us being in the room yet.
There is a similar requirement for knocking, where we will eventually do the same thing, and need an entry in the rooms table as well. Thus, reusing this function works, however its name needs to be generalised a bit.
Separated out from #6739.
The main use case is to see how many requests are being made, and how
many are second/third/etc attempts. If there are large number of retries
then that likely indicates a delivery problem.
If the script fails (or is CTRL-C'ed) between porting some of the events table and copying of the sequences then the port script will immediately die if run again due to the postgres DB having inconsistencies between sequences and tables.
The fix is to move the porting of sequences to before porting the tables, so that there is never a period where the Postgres DB is inconsistent. To do that we need to change how we port the sequences so that it calculates the values from the SQLite DB rather than the Postgres DB.
Fixes#8619
This should hopefully speed up `get_auth_chain_difference` a bit in the case of repeated state res on the same rooms.
`get_auth_chain_difference` does a breadth first walk of the auth graphs by repeatedly looking up events' auth events. Different state resolutions on the same room will end up doing a lot of the same event to auth events lookups, so by caching them we should speed things up in cases of repeated state resolutions on the same room.
`adbapi.ConnectionPool` let's you turn on auto reconnect of DB connections. This is off by default.
As far as I can tell if its not enabled dead connections never get removed from the pool.
Maybe helps #8574
* Check support room has only two users
* Create 8728.bugfix
* Update synapse/server_notices/server_notices_manager.py
Co-authored-by: Erik Johnston <erik@matrix.org>
Co-authored-by: Erik Johnston <erik@matrix.org>
If SSO login is used (e.g. SAML) in a multi worker setup, it should be mentioned that currently all SAML logins must run on the same worker, see https://github.com/matrix-org/synapse/issues/7530
Also, if you are using different ports (for example 443 and 8448) in a reverse proxy for client and federation, the path `/_matrix/media` on the client and federation port must point to the listener of the `media_repository` worker, otherwise you'll get a 404 on the federation port for the path `/_matrix/media`, if a remote server is trying to get the media object on federation port, see https://github.com/matrix-org/synapse/issues/8695
This PR adds some documentation that:
* Describes who the audience for the `docs/`, `docs/dev/` and `docs/admin/` directories are, as well as Synapse's wiki page.
* Stresses that we'd like all documentation to be down in markdown.
I idly noticed that these lists were out of sync with each other, causing us to miss a table in a test case (`local_invites`). Let's consolidate this list instead to prevent this from happening in the future.
This PR fixes two things:
* Corrects the copy/paste error of telling the client their displayname is wrong when they are submitting an `avatar_url`.
* Returns a `M_INVALID_PARAM` instead of `M_UNKNOWN` for non-str type parameters.
Reported by @t3chguy.
* Tie together matches_user_in_member_list and get_users_in_room
* changelog
* Remove type to fix mypy
* Add `on_invalidate` to the function signature in the hopes that may make things work well
* Remove **kwargs
* Update 8676.bugfix
* Tie together matches_user_in_member_list and get_users_in_room
* changelog
* Remove type to fix mypy
* Add `on_invalidate` to the function signature in the hopes that may make things work well
* Remove **kwargs
* Update 8676.bugfix
We do it this way round so that only the "owner" can delete the access token (i.e. `/logout/all` by the "owner" also deletes that token, but `/logout/all` by the "target user" doesn't).
A future PR will add an API for creating such a token.
When the target user and authenticated entity are different the `Processed request` log line will be logged with a: `{@admin:server as @bob:server} ...`. I'm not convinced by that format (especially since it adds spaces in there, making it harder to use `cut -d ' '` to chop off the start of log lines). Suggestions welcome.
Cached functions accept an `on_invalidate` function, which we failed to add to the type signature. It's rarely used in the files that we have typed, which is why we haven't noticed it before.
otherwise non-state events get written as `<FrozenEvent ... state_key='None'>`
which is indistinguishable from state events with the actual state_key `None`.
This modifies the configuration of structured logging to be usable from
the standard Python logging configuration.
This also separates the formatting of logs from the transport allowing
JSON logs to files or standard logs to sockets.
Not being able to serialise `frozendicts` is fragile, and it's annoying to have
to think about which serialiser you want. There's no real downside to
supporting frozendicts, so let's just have one json encoder.
I was trying to make it so that we didn't have to start a background task when handling RDATA, but that is a bigger job (due to all the code in `generic_worker`). However I still think not pulling the event from the DB may help reduce some DB usage due to replication, even if most workers will simply go and pull that event from the DB later anyway.
Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
This allows trailing commas in multi-line arg lists.
Minor, but we might as well keep our formatting current with regard to
our minimum supported Python version.
Signed-off-by: Dan Callahan <danc@element.io>
The test runner isn't present in the `[all]` set of extras, so the
previous instructions did not work without also installing `[test]`.
Note that this does not include the `[lint]` extras, since those do not
install on all supported Python versions (specifically, isort 5.x
requires Python 3.6, while we still support 3.5). Instructions for that
are included in our pull request template, so we should be fine there.
I've also dropped the `--no-use-pep517` arg to `pip install` since it
seems to have been added to address a temporary regression in pip 19.1
which was fixed in pip 19.1.1 the following month.
Lastly, updated the example output of the test suite to set more
realistic expectations around run time.
Signed-off-by: Dan Callahan <danc@element.io>
This is a requirement for [knocking](https://github.com/matrix-org/synapse/pull/6739), and is abstracting some code that was originally used by the invite flow. I'm separating it out into this PR as it's a fairly contained change.
For a bit of context: when you invite a user to a room, you send them [stripped state events](https://matrix.org/docs/spec/server_server/unstable#put-matrix-federation-v2-invite-roomid-eventid) as part of `invite_room_state`. This is so that their client can display useful information such as the room name and avatar. The same requirement applies to knocking, as it would be nice for clients to be able to display a list of rooms you've knocked on - room name and avatar included.
The reason we're sending membership events down as well is in the case that you are invited to a room that does not have an avatar or name set. In that case, the client should use the displayname/avatar of the inviter. That information is located in the inviter's membership event.
This is optional as knocks don't really have any user in the room to link up to. When you knock on a room, your knock is sent by you and inserted into the room. It wouldn't *really* make sense to show the avatar of a random user - plus it'd be a data leak. So I've opted not to send membership events to the client here. The UX on the client for when you knock on a room without a name/avatar is a separate problem.
In essence this is just moving some inline code to a reusable store method.
it seems to be possible that only one of them ends up to be cached.
when this was the case, the missing one was not fetched via federation,
and clients then failed to validate cross-signed devices.
Signed-off-by: Jonas Jelten <jj@sft.lol>
Split admin API for reported events in detail und list view.
API was introduced with #8217 in synapse v.1.21.0.
It makes the list (`GET /_synapse/admin/v1/event_reports`) less complex and provides a better overview.
The details can be queried with: `GET /_synapse/admin/v1/event_reports/<report_id>`.
It is similar to room and users API.
It is a kind of regression in `GET /_synapse/admin/v1/event_reports`. `event_json` was removed. But the api was introduced one version before and it is an admin API (not under spec).
Signed-off-by: Dirk Klimpel dirk@klimpel.org
==============================
Bugfixes
--------
- Fix bugs where ephemeral events were not sent to appservices. Broke in v1.22.0rc1. ([\#8648](https://github.com/matrix-org/synapse/issues/8648), [\#8656](https://github.com/matrix-org/synapse/issues/8656))
- Fix `user_daily_visits` table to not have duplicate rows per user/device due to multiple user agents. Broke in v1.22.0rc1. ([\#8654](https://github.com/matrix-org/synapse/issues/8654))
-----BEGIN PGP SIGNATURE-----
iQFEBAABCgAuFiEEBTGR3/RnAzBGUif3pULk7RsPrAkFAl+W6QIQHGVyaWtAbWF0
cml4Lm9yZwAKCRClQuTtGw+sCR/EB/9Afs/QiqD7uJ7XtAASwlOxcpSrvB6xSMzU
VIiYLrUSCvHe8SwcPG6nToy69hHmwLBPu9e8gvlQkOY4nJpW9VTGvhFcP5eCgeKd
vDp+elVD6BPg+6S3RaI/F5hyQ7gn9qZ/WexkVixDkVs5eAL3gNlZcjw9RwaT9ei9
fEOYlMi1CRSRzKVBPMEFCAvOpJJ8rA/iq4JK7+P0TplkhD+HydG0cYr6OoL7CmaD
L/au565Ynp6o0hSpNpTFqW6C5W5JRvz/TyG9/N95h9+WqubZzsPPDUzh98B3eBoX
43Wu9Qoc3mNamf0++HbHGUPI8dRV1gssxL01hs/iLUZz7CxXY+Ir
=Z3sQ
-----END PGP SIGNATURE-----
Merge tag 'v1.22.0rc2' into develop
Synapse 1.22.0rc2 (2020-10-26)
==============================
Bugfixes
--------
- Fix bugs where ephemeral events were not sent to appservices. Broke in v1.22.0rc1. ([\#8648](https://github.com/matrix-org/synapse/issues/8648), [\#8656](https://github.com/matrix-org/synapse/issues/8656))
- Fix `user_daily_visits` table to not have duplicate rows per user/device due to multiple user agents. Broke in v1.22.0rc1. ([\#8654](https://github.com/matrix-org/synapse/issues/8654))
* Fix user_daily_visits to not have duplicate rows for UA.
Fixes#8641.
* Newsfile
* Fix typo.
Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
#8567 started a span for every background process. This is good as it means all Synapse code that gets run should be in a span (unless in the sentinel logging context), but it means we generate about 15x the number of spans as we did previously.
This PR attempts to reduce that number by a) not starting one for send commands to Redis, and b) deferring starting background processes until after we're sure they're necessary.
I don't really know how much this will help.