CRITICAL - Socket Timeout

marshalleq

Dude, you haven’t read properly. a) The disk issues are on different disks on a different controller. I have a whole other thread about this over on truenas if you want to fault find that for me. 😉. If you know ZFS you know it actually tells you if there are issues with data too and names any corrupted files, which the mail server doesn’t have any / never had any because it’s disks are not having issues. b) And answering your final idea, I have actually completely backed this up, restored it to a new host, which is also mentioned above. Where did you get the idea I don’t have backups?

SimplyCorbett

marshalleq But if you have backups from a working copy they should also work when restored. So.. you may want to be raising an issue on github if you feel its a problem with a mailcow update.

marshalleq

Where did I say it was a problem with a mailcow update? All I’m posting here for is to see if anyone knows what these logs are saying and if they’ve come across it / can offer help. Github is generally not a support forum, so I’d want to be confident it is actually a code bug before posting there.

SimplyCorbett

Honestly if you are restoring known good working backups and running into this issue and it is not a mailcow update causing it then the only other thing I can think of is you have a unstable system.

Worker process died unexpectedly

Does not fill me with confidence that your system is entirely stable.

marshalleq

OK thanks, in your experience, what causes Worker Process died unexpectedly?

esackbauer

I had once a weird issue on my firewall, where processes kept crashing and producing core dumps.
No indication in the log files of any hardware error or related issue.
After a lot of troubleshooting it turned out a be a defective memory module. My firewall was not using ECC memory.
I suggest your server has ECC memory installed? anything in the BIOS logs regarding memory errors?

marshalleq

I have done extensive memory testing on the box due to the disk issues. No memory errors showing at all. Some months back I switched servers and this new one doesn’t have ECC, it could be that, but it’s unlikely given the extensive memory testing I think. However, I may just look at getting some. Might be a bit cheaper given it’s DDR-4 and we’re at DDR-5 now I think.

SimplyCorbett

I dont think you ever said. How many active clients are using your server?

marshalleq

Good point, I didn’t say clients, but I did say 3 users. I’ve doubled the active Sogo workers anyway. In effect, me as one user I have maybe 5 mailboxes plus other the other 2 users have maybe 4 so I guess calling it 10 users and I have 3 devices, plus they will have a few too so perhaps that makes it like 30 users?

SimplyCorbett

Well, what happens if you restore your backup on some cheap $5 VPS from say hetzner?

Does it work as expected?

If it does not, try to restore from a -earlier- backup. And keep doing this until you find one that works.

At which point my assumption would be something in the backup corrupted during your hardware issues early on.

To clarify how are you restoring your backups and when you backed up your mailcow instance did you backup all?

You are restoring backups from the mailcow backup script and not ZFS snapshots, correct?

marshalleq

Assuming you’re replying to the last questions you asked, your reply makes no sense to me. I don’t understand the relationship between numbers of users and you giving me advice on backups. Did you mean to post this somewhere else? Are you just trying to get your moo level up or something?

SimplyCorbett

marshalleq Nah, I was asking about users because I wanted to know if you were running into resource exhaustion. With 3 users and 16GB RAM on the VM thats unlikely.