Conversation:
Notices
-
@takeshitakenji Sounds like PHP-FPM is acting up, it's the only thing being gated by nginx. Check memory use of the php things. I suspect the redis queue may be causing you trouble trying to force stuff with retries.
-
@takeshitakenji Unless you have some sort of weird setup with nginx anyways.
-
@takeshitakenji Keep in mind that you can have error states with a non-error situation if it hangs, or takes a long time to succeed. So it might not technically be in an error state, but it can effectively be.
-
@takeshitakenji Could be that they're just failing after the normal retries like normal. What exactly is the wonkiness you're seeing?
-
@takeshitakenji Probably locks from the queue_item
-
@takeshitakenji To explain further the DBO object will lock the rows related to a queue item while the queue item is being processed so if the message you're trying to delete is still federating, it will block a deletion. This is probably not desirable behaviour, because we'd want to stop it BEFORE federation if you delete it, but I'm not describing the ideal here, …
-
@takeshitakenji I mean obviously in most cases we CANT delete a message before it federates, but if we have the opprotunity to then we should take it IMO
-
@takeshitakenji To trace up that metaphorical stack some, if it detects a deletion for that same notice is already in the queue, just unprocessed, it won't duplicate that deletion.
-
@takeshitakenji heh I'm trying. This condition is kicking my ass the last couple days though