Recent searches
Search options
Mastodon tech question
@puzzled nope, just people who are wondering the same thing boosting my toot lol.
I don't get *too* many spammer signups, so I've just been suspending them through the web admin interface, which is what generates all those sidekiq tasks and caused the most recent service disruptions. I've tuned it to dual back concurrency and am not deleting any accounts right now and it's been running very well since (no errors at all for 24 hours, yay!), so I'll leave it alone for now...
Mastodon tech question
@puzzled ...there is also a delete user "tootctl" command I can execute from a shell which I'll try early Saturday morning (like 2am early) on one of the unconfirmed users. If that method also does the same thing I'll raise an issue on the Mastodon github.
But aside from that expect stability to improve a lot...it was very solid for 2+ years before recent OS upgrades and performance tweaks!
Mastodon tech question
Well @puzzled I'm still puzzled too lol.
The tootctl commands don't do anything special--deleting unconfirmed accounts with zero activity from the command line also spawns thousands of jobs per user in the queue too, and still nobody can explain why this is necessary.
My system is still stable under normal use, so I will go back to looking at why sidekiq crashes when the queue is flooded before I do any more user deletes. Might still be how docker mounts storage volumes.