Issue Plesk freezes every 1-2 days

[JD165]

Server operating system version

AlmaLinux 8.9 (Midnight Oncilla)

Plesk version and microupdate number

18.0.59 Update #2

Goodmorning,

We have a plesk server running with some domains on it. Every 1-2 days plesk randomly freezes and wont come back untill i reboot de VPS itself.
Monitoring shows constant 2-10% cpu and 30% memory usage, no strange behavior in that.

This morning we hade the same issue and on the VPS console i saw the errors, i've attached them in this post.

Could this be connected with the freeze issue? Any idea what else it could be that is causing this(maybe its a known bug)?
We have this VPS for 4 months now and the issues started about 1 month ago and the interval between the freezes are increasing.

Thanks in advance.

 

[Peter Debik]

The screenshot only shows that tasks are hanging. There should also be a reason given. Example:

INFO: task sendmail:14350 blocked for more than 120 seconds.
"echo 0 /proc/sys/kernel/hung_task_timeout_secs" disables this message.
sendmail D ffff810001004320 0 14349 16022 14355 12937 (NOTLB)
ffff81001b72bdc8 0000000000000086 ffff8201d91be700 ffffffff800099ae
ffff8101d8acf0c0 0000000000000009 ffff8200172527a0 ffffffff80309b60
0010f4837aa8f02a 0000000000004646 ffff820017252988 0000000006a5ca98
Call Trace:
[ffffffff800099ae] __link_path_walk+0x173/0xf42
[ffffffff8002cd2c] mntput_no_expire+0x19/0x89
[ffffffff80064c6f] __mutex_lock_slowpath+0x60/0x9b
[ffffffff800236d7] __path_lookup_intent_open+0x56/0x97
[ffffffff80064cb9] .text.lock.mutex+0xf/0x14
[ffffffff8001afe7] open_namei+0xea/0x6d5
[ffffffff800274fb] do_filp_open+0x1c/0x38
[ffffffff80019e1e] do_sys_open+0x44/0xbe
[ffffffff8005e28d] tracesys+0xd5/0xe0

This could look different, but there should be such a part in your syslog, too that points to the reason of the issue.

Frequently unresponsive disks like NFS can cause it. Other reasons can be hardware issues.

 

[JD165]

Hi,

It just happened again. Plesk was reachable but none of the websites were, I restarted apache trough Plesk and it started working again.

In the log messages i see this: AH03490: scoreboard is full, not at MaxRequestWorkers.Increase ServerLimit

Is it possible that this is related to the crashing?

Kr Justin

 

[AYamshanov]

Agreed with Peter. If you checked Monitoring graphs, have you checked disks graphs?
If I understood correctly, this server is kind of VPS; do you have ability to diagnose the host system for any issues? On the host system, is there any limitations for the guest VMs?

 

[JD165]

Hi,

Yes the plesk is running on a VPS. I've diagnosed the VM itself with the vendor and all seems good/great, its at 5-10% all the time, no peaking.

pm.max_requests is on 0, could that be connected with the issue?

 

[Peter Debik]

AH03490: scoreboard is full, not at MaxRequestWorkers.Increase ServerLimit

Your Apache is probably using MPM-Event mode. Instead, try using MPM-Prefork. It can be selected in the Tools & Settings > Apache dialog.

 

[JD165]

Your Apache is probably using MPM-Event mode. Instead, try using MPM-Prefork. It can be selected in the Tools & Settings > Apache dialog.

Hi,

We changed it to prefork mode, seemed to be going good but just now we had the same issue again. Plesk is available at that time but none of the websites that are hosted on it.
When I restart apache from the manage services within plesk, all the sites start to work again.

 

[JD165]

Extra information, looks like the crashing comes together with the following error:

AH00161: server reached MaxRequestWorkers setting, consider raising the MaxRequestWorkers setting

 

[JD165]

Sorry for the spam, but would it help if i define pm.max_requests?