We use retry_event in queue_processors.py to handle trying on failures,
without getting stuck in permanent retry loops if the event ends up
leading to failure on every attempt and we just keep sending NACK to
rabbitmq forever (or until the channel crashes). Tornado queues haven't
been using this, but they should.
setup_event_queue() generates some logs about loaded event queues, and
it's good for the logging system to have access to the port at that
point already.
Zulip has had a small use of WebSockets (specifically, for the code
path of sending messages, via the webapp only) since ~2013. We
originally added this use of WebSockets in the hope that the latency
benefits of doing so would allow us to avoid implementing a markdown
local echo; they were not. Further, HTTP/2 may have eliminated the
latency difference we hoped to exploit by using WebSockets in any
case.
While we’d originally imagined using WebSockets for other endpoints,
there was never a good justification for moving more components to the
WebSockets system.
This WebSockets code path had a lot of downsides/complexity,
including:
* The messy hack involving constructing an emulated request object to
hook into doing Django requests.
* The `message_senders` queue processor system, which increases RAM
needs and must be provisioned independently from the rest of the
server).
* A duplicate check_send_receive_time Nagios test specific to
WebSockets.
* The requirement for users to have their firewalls/NATs allow
WebSocket connections, and a setting to disable them for networks
where WebSockets don’t work.
* Dependencies on the SockJS family of libraries, which has at times
been poorly maintained, and periodically throws random JavaScript
exceptions in our production environments without a deep enough
traceback to effectively investigate.
* A total of about 1600 lines of our code related to the feature.
* Increased load on the Tornado system, especially around a Zulip
server restart, and especially for large installations like
zulipchat.com, resulting in extra delay before messages can be sent
again.
As detailed in
https://github.com/zulip/zulip/pull/12862#issuecomment-536152397, it
appears that removing WebSockets moderately increases the time it
takes for the `send_message` API query to return from the server, but
does not significantly change the time between when a message is sent
and when it is received by clients. We don’t understand the reason
for that change (suggesting the possibility of a measurement error),
and even if it is a real change, we consider that potential small
latency regression to be acceptable.
If we later want WebSockets, we’ll likely want to just use Django
Channels.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This is a preparatory commit for using isort for sorting all of our
imports, merging changes to files where we can easily review the
changes as something we're happy with.
These are also files with relatively little active development, which
means we don't expect much merge conflict risk from these changes.
Previous cleanups (mostly the removals of Python __future__ imports)
were done in a way that introduced leading newlines. Delete leading
newlines from all files, except static/assets/zulip-emoji/NOTICE,
which is a verbatim copy of the Apache 2.0 license.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This should make it possible for there to safely be multiple Tornado
processes running on different ports on the same system.
It may also fix a rare race bug in development, where previously, it
was possible for the Tornados processes for Casper and the main
development server to interfere; I haven't investigated whether this
was a real bug or not, but now those two services will use independent
Tornado files.
We still need to add something to direct traffic between the different
Tornado processes.
This logging was apparently broken when sorting imports; it's a fairly
unique thing in our codebase that this would be a problem. Prevent
future regressions by adding this exception explicitly to the isort
configuration.
runtornado unbuffers its output using
sys.stdout = os.fdopen(sys.stdout.fileno(), 'w', 0).
This is not python 3 compatible since we can't specify
buffering on a text stream in python 3. So use the '-u'
option of python when calling runtornado.py to make output
unbuffered.
The new Tornado handler tracking logic properly handled requests that
threw an exception or followed the RespondAsynchronously code path,
but did not properly de-allocated the handler in the syncronous case.
An easy reproducer for this is to load a new Zulip browser window;
that will leak 2 handler objects for the 2 synchronous requests made
from Django to Tornado as part of initial state fetching.
The recent Tornado memory leak fix
(1396eb7022) didn't use the correct
variable name for the current handler ID, causing this cleanup code to
fail in the event that a view raised an exception.
In 2ea0daab19, handlers were moved to
being tracked via the handlers_by_id dict, but nothing cleared this
dict, resulting in every handler object being leaked. Since a Tornado
process uses a different handler object for every request, this
resulted in a significant memory leak. We fix this by clearing the
handlers_by_id dict in the two code paths that would result in a
Tornado handler being de-allocated: the exception codepath and the
handler disconnect codepath.
Fixes#463.
Previously, client descriptors were referenced directly from the
handler object. Once we split the Tornado process into separate queue
and connection servers, these will no longer be in the same process,
so we need to reference them by ID instead.
Mateusz Mandera
Anders Kaseorg
Tim Abbott
neiljp (Neil Pilgrim)
rht
Umair Khan
nikolay
Rafid Aslam
K.Kanakhin
Eklavya Sharma
Tim Abbott
Varshit
Tim Abbott