Log all unknown IRC events to figure out the IRC ghosts #40
Conversation
| // noise in the logfile but will hopefully tell me which | ||
| // command creates the IRC "ghosts" (no longer online IRC users, | ||
| // that are still listed on the metaserver) | ||
| switch e.Code { |
There was a problem hiding this comment.
Have you tried how much data this produces? I do not know what sort of events we receive here, but if this is for everything that happens on IRC, could this overflow our disk on the server realistically?
There was a problem hiding this comment.
I have no idea and I don't know much about IRC anyway. I don't think it will request any messages we wouldn't receive anyway, so the amount of network traffic exchanged with the IRC server should be an upper limit. Based on that, I doubt that it can fill the disk, but it should be under observation while this is deployed.
Since the bug report describes a way to trigger the "ghost bug", I was considering deploying this branch, asking for someone to change the nickname a few times and hopefully create ghosts that way. Then I could look at the logs which IRC messages have to be handled additionally to avoid them.
There was a problem hiding this comment.
the repro plan sounds great. Also deploying and watching the disk space is fine. Just make sure that we are not flooding the disk, because a system with full disk is nearly unfixable - i.e. hard to ssh into and do anything.
There was a problem hiding this comment.
I deployed this branch two days ago and until now this logging didn't made any problems. It was logging a few PING/PONG messages (that I added to the ignore-list) and it is logging some messages when connecting to the IRC server, but nothing else yet. So until now nothing that fills the disk.
Co-Authored-By: Notabilis <[email protected]>
Issue #37 provided the clue that renaming oneself on IRC leaves a "ghost" with the old name in the lobby. Since I wasn't able to find out the protocol command for renaming, let's simply log all received IRC commands until it happens again.
If someone is able to provide this information, that would be fine as well.