How to kick out stalled sessions?

In one of my conference rooms there are several stalled session, i.e. there are users, who obviously left the session without properly closing it (hitting the red button). One of them has still the rights of the owner, so those rights are no longer usable for anyone.
Is there a way to stop such stalled sessions, getting the room free for another (new) use?

I am also experiencing this problem as well, tried to change the bosh_max_inactivity to 1 (second), as per https://prosody.im/doc/modules/mod_bosh , but it seems the ghost user still linger a while before being auto kicked out by the Jitsi server.

Changing bosh session timeout is not a good idea. Which prosody version is this?

Currently running this version from apt (on Ubuntu 18.04).
prosody /bionic,now 0.10.0-1build1 amd64 [installed,automatic]

Any alternative suggestion, we need to get rid of the Ghost participant as soon as possible otherwise users are asking why there is an extra person online.

upgrade to prosody 0.11 (don’t forget to change storage setting in jitsi-meet prosody config). Also enable epoll network backend in prosody, this is an optimisation for performance.

thank you, will try that.

Hi, just a quick update, initial test does seems like we are able to kick out of app crashed (simulated via killing jitsi app) almost immediately! Thank you.

Hi,

Did you solve this problem? Could you share the steps to enable epoll network backend in prosody?

Thanks

Kind of, if disconnection is proper, then no issue, but if disconnection due to app crashed, then it still take a while before the ghost user disappear. Prosody is standard setup I presume (our team member installed it), but we did upgrade to prosody 0.11.5.

you can refer to this link https://prosody.im/doc/network_backend#epoll

Hi,

Thanks for the info, do you also upgrade prosody version?

yup, current latest stable: 0.11.5

is there any method for upgrading prosody on a live server without disrupting the service?

Also curious on if it is possible to upgrade on a live server without disrupting the service.

That is not possible when using a single machine. You need a way to have several shards, install new one, test them and rotate them.