mysql server crashes every 18 days, no oom, no crash logs, no sigkill being sent (used auditd). so I figure it's a unknown corner case bug in mysql. now I use a cron job to restart the damn thing every week at 3am, not a problem anymore

Why not migrate db to a new host

Every 18 days at the same time?

@donkeyScript 18 days around 11pm-3am cst

Hmm, sounds like a cron job was created to kill the server every 3 weeks.

@NullBot sounds like it but I checked for it, nothing that messes with mysql. besides a weekly restart wouldn't fix that. my first thought was that it was running out of open files but I monitored that also and got nowhere. but judging by my restart their might be a problem in the cleanup somewhere since 11pm-3am is the quite time on the server

@jckimble Since I had a similar problem on a previous job of mine, are you sure there isn't an overly eager cleaning lady that unplugs everything to wipe the sockets?

@Godisalie same thing about the cron fixing the problem, and if the kernel was sending a sigkill auditd would have caught it

@jckimble I do like a good mystery.