Hi,
I had a very peculiar incident over the Easter period with my self-hosted Taiga.
My server essentially used up all its ‘open files’ limit and I needed to reboot to recover.
Analysing my logs, I identified that there was a steeply growing number of new established connections from IP addresses known to use my Taiga instance.
They were all in the ‘ESTABLISHED’ tcp state and port 443, I suspect that the browser was continually making new websocket connections and yet not releasing the old. What is peculiar is that this wasn’t reflected in the Nginx logs themselves…
See attached image, you can see where the ‘runaway’ number of connections occurred.
I realise this is a vague query - and I can mitigate it with periodic restarts if it comes up again (as well as raising the ulimit for open files), but I am mostly interested to hear if anyone else has ever experienced, with Taiga, a ‘runaway’ scenario with a growing number of websocket connections. Has it ever occurred for the ‘PaaS’ hosted Taiga Tree, for example?
Thanks!
