Issue Apache just stop working for hours with 0% CPU usage

[Tiria]

Server operating system version

Ubuntu 18.04.6 LTS

Plesk version and microupdate number

18.0.49 update 2

Hello there,

Sometimes, but always at night around 2:00-3:00 am and for some hours, the apache process just stop responding, leaving all websites with a nginx 504 error (proxy timeout, almost all website are configured with nginx+apache).
Apache service is still running on that period but is just not responding (see chart attached).

I tried to find errors in sites and plesk logs but dont find anything which could match this error. All apache logs are empty on the period and all error found just before is some php warning that always occurring.

There is no cron around 2:00-3:00 am, just some backups (all backups are programmed along the night).

Is someone have an idea to check why this problem occur ?

Thanks

 

[wiredhosting]

Hi. I Would try two different things.

1) try disabling backups. Just in case.(remember after 24hs to re enable them)

2) try to swith to only apache instead of apache+nGinx.

After 24hs check the graphs and let us Knowles if anyrhing changed.

 

[Maarten]

The Plesk backup process can be rather CPU-consuming.
Try running the backup with "no compression" and at priority 19:

 

[Tiria]

Thanks for your answers.

Hi. I Would try two different things.

1) try disabling backups. Just in case.(remember after 24hs to re enable them)

2) try to swith to only apache instead of apache+nGinx.

After 24hs check the graphs and let us Knowles if anyrhing changed.

The problem is (perhaps it was ambiguous in my previous post) that the problem does not happen every day but only few times in the month (but every time it happens it is in the same hours). I can't just disable backup waiting for the bug to happen again.

The Plesk backup process can be rather CPU-consuming.
Try running the backup with "no compression" and at priority 19:

I'll try to decrease priority level to 19 (It was set up to 10) and lower simultaneous backups to 4 instead of 10. But I think I can't disable compression because of space available in our backup server. Is there a way to set up different compression level per domains (some of them are too large for the compression to be disable) ?

That being said, we already tried to change backup scheduling and this has not changed the times at which the problem occurs, so I'm not sure it was backup related.

 

[mow]

Have you checked the timestamps of the backup files to verify they're not in that time range?