On Saturday, February 18, 2023 at 7:50:56 AM UTC-6 Damian Minkov wrote: Hi,
On Sat, Feb 18, 2023, 04:59 Matthew Wild <[email protected]> wrote: Hi Damian, On Fri, 17 Feb 2023 at 13:58, Damian Minkov <[email protected]> wrote: > > We have been running this patch for more than a month without noticing any major issues ... one thing I noticed today is this error, have seen it only once: Sorry for the delayed response on this thread. Past few months have been busy. It sounds like the original issue is easy to reproduce for you. Any chance you can try it with the latest 0.12 code (e.g. the 'prosody-0.12' package from our repo)? We fixed a bug recently related to smacks hibernation, and I'm not sure whether this is related or not. I will check it out next week and let you know. This is different than 0.12.2? I see now the various session destruction commits in the 0.12 branch after 0.12.2. I will test it first thing on Monday and will let you know. Thanks Skimming your patch, I'm not clear on exactly what the root cause is. A hibernating session wouldn't actually be destroyed (just session.conn removed), so I would expect things like the outgoing stanza queue to remain. Actually, I'm also not entirely clear on what the issue you're experiencing is. The "hibernating too long" message is not necessarily an error - it's expected behaviour if the client disconnects uncleanly and does not return before the configured hibernation timeout, which can be very common. In the case of a clean disconnect, the hibernation code should not kick in at all. In the case of hibernation, the session should not be destroyed until after the timeout period. Debug logs during the disconnect of a client would help identify which of these is happening. Yep, we saw the problem as we added some logs to the unbind event ( https://github.com/jitsi/infra-configuration/blob/ae1b579bce46cfbe686ec2cfc4b44b6e2b0309d2/ansible/roles/prosody/files/mod_muc_webhooks.lua#L706). To detect those sessions that actually lost connection. And then we noticed that we got the print for hibernation for all sessions even for the cleanly disconnected, the print is of course a minute after the clean disconnect (this is our setting for hibernating a connection duration). The order of calls is wrong and on clean disconnect, sessions are not cleared correctly and all go into hibernation. Add the unbind message and do a clean connect and disconnect via the websocket and you will able to repro. I hope this makes more sense. Thanks damencho Regards, Matthew -- You received this message because you are subscribed to the Google Groups "prosody-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/prosody-dev/CAJt9-x7t2nMjYBopCTRAG%3DiVMt2RZqsr_Pm6MhBnO-ZrhRb_rQ%40mail.gmail.com . -- You received this message because you are subscribed to the Google Groups "prosody-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/prosody-dev/269d0c66-5690-4646-8d6d-5319d0815a94n%40googlegroups.com.
