I suspect #footiMac is getting a little overtaxed. But I am wondering if other #selfhosted#mastoAdmin have a thought about these messages in Sidekiq for dead processes? Too many files open, on my end or the remote end?
Is there maybe a FAQ available for small admins like me to look to that might show the most common indicators of problems on a Mastodon host?
I remain concerned that Java/Elastisearch is putting more load on my server than is really warranted. Thoughts? #help#MastoHelp#Mastodon#SelfHost
I took the example from the resource linked before. This was my procedure:
I increased the sidekiq process to allow 75 instead of the default 25 workers. This helped, but it was still full and only going down slowly.
Then I split out the processes (by duplicating the systemctl file and make one each for the ingress, push+pull, default, scheduler, and mailers. All had 75 workers. It turned out only the pushpull service was pinned to that 75 and database connections went to over 200.
So I split the push and pull into separate processes and realized it was all on the pull process. All other processes were very low.
Finally, I reduced the number of workers back down to 25 for all processes except the pull process. RAM usage has quadrupled from 400MB to 1700 but I had space there to grow into.
My server is now back to a good place, database connections are below 50, nothing in the queue, and toots/notifications are going out instantly again.
And most importantly, I can once again handle all the news GottaLaff puts out! :D