-
-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Unhandled Error: doWrite called on a twisted.internet.udp.Port #4727
Comments
I'm also still seeing this error with 1.1.0
|
It's also the last thing I see in my logs before my homeserver stops working :( I'm not sure it's related. |
are you using a network log server for your log output? |
Nope.
Sent from Nine
…________________________________
From: Richard van der Hoff <notifications@github.com>
Sent: Monday, July 8, 2019 03:14
To: matrix-org/synapse
Cc: David Kowis; Comment
Subject: Re: [matrix-org/synapse] Unhandled Error: doWrite called on a twisted.internet.udp.Port (#4727)
are you using a network log server for your log output?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or mute the thread.
|
I have a similar error on synapse 1.20.1:
we have a federation problem between two of our homeservers: If you send from one, it is seen on the other immediately, but vica-versa not. only if you interact with a user from a third server, the message arrive at that later tie then (but with the original timestamp). |
Similar on Synapse v1.37.1. Logging configuration:
|
@oleg-fiksel can you share the stacktrace? A fair bit has changed since 1.20.1. |
|
I'm at a loss to explain this. It seems to be coming from the depths of Twisted. |
We solved our federation problem by adding a redirect to the .well-known path: There was a redirect (301) when calling https://matrix.eclabs.de/.well-known/matrix/server – this redirect ended up in nirvana and returned a 404. We added this to our nginx conf:
Maybe this was related? |
@rubo77 to be clear: you saw this exception when I'm struggling to imagine how changing your own .well-known file can fix errors on your own server, but at this point I'm not ruling anything out... |
Exactly |
I'm not able to reproduce it by setting |
https://twistedmatrix.com/trac/ticket/4759 looks to be a possibly related ticket with the same stack trace, which suggests the problem might be that we end up doing things with the reactor before daemonizing? I'm not sure though, or how to progress from here |
Seems to show up on federation_sender workers when they have a lot of work to do (eg new login for a user with many rooms) |
I've saw this message in my error log:
Version information
The text was updated successfully, but these errors were encountered: