Propagate opentracing contexts across workers
Also includes some Convenience modifications to opentracing for servlets, notably:
- Add boolean to skip the whitelisting check on inject
extract methods. - useful when injecting into carriers
locally. Otherwise we'd always have to include our
own servername and whitelist our servername
- start_active_span_from_request instead of header
- Add boolean to decide whether to extract context
from a request to a servlet
Add authenticated_entity and servlet_names tags.
Functionally:
- Add a tag for authenticated_entity
- Add a tag for servlet_names
Stylistically:
Moved to importing methods directly from opentracing.
This refactors MatrixFederationAgent to move the SRV lookup into the
endpoint code, this has two benefits:
1. Its easier to retry different host/ports in the same way as
HostnameEndpoint.
2. We avoid SRV lookups if we have a free connection in the pool
If we have recently seen a valid well-known for a domain we want to
retry on (non-final) errors a few times, to handle temporary blips in
networking/etc.
This gives a bit of a grace period where we can attempt to refetch a
remote `well-known`, while still using the cached result if that fails.
Hopefully this will make the well-known resolution a bit more torelant
of failures, rather than it immediately treating failures as "no result"
and caching that for an hour.
It costs both us and the remote server for us to fetch the well known
for every single request we send, so we add a minimum cache period. This
is set to 5m so that we still honour the basic premise of "refetch
frequently".
--------
- Fix a regression introduced in v1.2.0rc1 which led to incorrect labels on some prometheus metrics. ([\#5734](https://github.com/matrix-org/synapse/issues/5734))
-----BEGIN PGP SIGNATURE-----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=+JBn
-----END PGP SIGNATURE-----
Merge tag 'v1.2.0rc2' into develop
Bugfixes
--------
- Fix a regression introduced in v1.2.0rc1 which led to incorrect labels on some prometheus metrics. ([\#5734](https://github.com/matrix-org/synapse/issues/5734))
* Fix servlet metric names
Co-Authored-By: Richard van der Hoff <1389908+richvdh@users.noreply.github.com>
* Remove redundant check
* Cover all return paths
* Configure and initialise tracer
Includes config options for the tracer and sets up JaegerClient.
* Scope manager using LogContexts
We piggy-back our tracer scopes by using log context.
The current log context gives us the current scope. If new scope is
created we create a stack of scopes in the context.
* jaeger is a dependency now
* Carrier inject and extraction for Twisted Headers
* Trace federation requests on the way in and out.
The span is created in _started_processing and closed in
_finished_processing because we need a meaningful log context.
* Create logcontext for new scope.
Instead of having a stack of scopes in a logcontext we create a new
context for a new scope if the current logcontext already has a scope.
* Remove scope from logcontext if logcontext is top level
* Disable tracer if not configured
* typo
* Remove dependence on jaeger internals
* bools
* Set service name
* :Explicitely state that the tracer is disabled
* Black is the new black
* Newsfile
* Code style
* Use the new config setup.
* Generate config.
* Copyright
* Rename config to opentracing
* Remove user whitelisting
* Empty whitelist by default
* User ConfigError instead of RuntimeError
* Use isinstance
* Use tag constants for opentracing.
* Remove debug comment and no need to explicitely record error
* Two errors a "s(c)entry"
* Docstrings!
* Remove debugging brainslip
* Homeserver Whitlisting
* Better opentracing config comment
* linting
* Inclue worker name in service_name
* Make opentracing an optional dependency
* Neater config retreival
* Clean up dummy tags
* Instantiate tracing as object instead of global class
* Inlcude opentracing as a homeserver member.
* Thread opentracing to the request level
* Reference opetnracing through hs
* Instantiate dummy opentracin g for tests.
* About to revert, just keeping the unfinished changes just in case
* Revert back to global state, commit number:
9ce4a3d906
* Use class level methods in tracerutils
* Start and stop requests spans in a place where we
have access to the authenticated entity
* Seen it, isort it
* Make sure to close the active span.
* I'm getting black and blue from this.
* Logger formatting
Co-Authored-By: Erik Johnston <erik@matrix.org>
* Outdated comment
* Import opentracing at the top
* Return a contextmanager
* Start tracing client requests from the servlet
* Return noop context manager if not tracing
* Explicitely say that these are federation requests
* Include servlet name in client requests
* Use context manager
* Move opentracing to logging/
* Seen it, isort it again!
* Ignore twisted return exceptions on context exit
* Escape the scope
* Scopes should be entered to make them useful.
* Nicer decorator names
* Just one init, init?
* Don't need to close something that isn't open
* Docs make you smarter
Prevents a SynapseError being raised inside of a IResolutionReceiver and instead opts to just return 0 results. This thus means that we have to lump a failed lookup and a blacklisted lookup together with the same error message, but the substitute should be generic enough to cover both cases.
Firstly, we always logged that the request was being handled via
`JsonResource._async_render`, so we change that to use the servlet name
we add to the request.
Secondly, we pass the exception information to the logger rather than
formatting it manually. This makes it consistent with other exception
logging, allwoing logging hooks and formatters to access the exception
information.
Turns out that the library does a better job of parsing URIs than our
reinvented wheel. Who knew.
There are two things going on here. The first is that, unlike
parse_server_name, URI.fromBytes will strip off square brackets from IPv6
literals, which means that it is valid input to ClientTLSOptionsFactory and
HostnameEndpoint.
The second is that we stay in `bytes` throughout (except for the argument to
ClientTLSOptionsFactory), which avoids the weirdness of (sometimes) ending up
with idna-encoded values being held in `unicode` variables. TBH it probably
would have been ok but it made the tests fragile.
* Remove redundant WrappedConnection
The matrix federation client uses an HTTP connection pool, which times out its
idle HTTP connections, so there is no need for any of this business.
* Correctly retry and back off if we get a HTTPerror response
* Refactor request sending to have better excpetions
MatrixFederationHttpClient blindly reraised exceptions to the caller
without differentiating "expected" failures (e.g. connection timeouts
etc) versus more severe problems (e.g. programming errors).
This commit adds a RequestSendFailed exception that is raised when
"expected" failures happen, allowing the TransactionQueue to log them as
warnings while allowing us to log other exceptions as actual exceptions.
- Improve logging: log things in the right order, include destination and txids
in all log lines, don't log successful responses twice
- Fix the docstring on TransportLayerClient.send_transaction
- Don't use treq.request, which is overcomplicated for our purposes: just use a
twisted.web.client.Agent.
- simplify the logic for setting up the bodyProducer
- fix bytes/str confusions
If a connection is lost before a request is read from Request, Twisted
sets `method` (and `uri`) attributes to dummy values. These dummy values
have incorrect types (i.e. they're not bytes), and so things like
`__repr__` would raise an exception.
To fix this we had a helper method to return the method with a
consistent type.
If a HTTP handler throws an exception while processing a request we
automatically write a JSON error response. If the handler had already
started writing a response twisted throws an exception.
We should check for this case and simple abort the connection if there
was an error after the response had started being written.
The existing deferred timeout helper function (and the one into twisted)
suffer from a bug when a deferred's canceller throws an exception, #3842.
The new helper function doesn't suffer from this problem.
We want to wait until we have read the response body before we log the request
as complete, otherwise a confusing thing happens where the request appears to
have completed, but we later fail it.
To do this, we factor the salient details of a request out to a separate
object, which can then keep track of the txn_id, so that it can be logged.
The problem with dumping all of the json response into the Request object at
once is that doing so starts the timeout for the next request to be received:
so if it takes longer than 60s to stream back the response to the client, the
client never gets it.
The correct solution is to use a Producer; then the timeout is only started
once all of the content is sent over the TCP connection.
Erik Johnston