Issue sw-engine eating cpu

[vulcain]

Since the last update plesk onyx 17.5.3 there is a PHP-FPM settings. at first it was setup by default as "static" and it was eating all memory, I switched all sites to "ondemand" and now the server is down 6h a day.
I have to manually re-start services or reboot. The same server was working fine before this update.

I have no logs, no backup, no mobile extension and no scheduled task apart the ones from plesk.
I receive this kind of message 5 times a day from the server health. Anyone?

top - 05:45:29 up 7:44, 0 users, load average: 1.36, 32.33, 37.82
Tasks: 163 total, 5 running, 158 sleeping, 0 stopped, 0 zombie
%Cpu(s): 74.3 us, 22.9 sy, 0.0 ni, 0.0 id, 0.0 wa, 0.0 hi, 2.9 si, 0.0 st
KiB Mem : 3881672 total, 104716 free, 2768212 used, 1008744 buff/cache
KiB Swap: 0 total, 0 free, 0 used. 614916 avail Mem

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
8520 psaadm 20 0 373292 32940 19468 R 52.9 0.8 0:00.13 sw-engine
8518 psaadm 20 0 371476 30712 19544 R 41.2 0.8 0:00.10 sw-engine
8519 psaadm 20 0 371476 30708 19544 R 41.2 0.8 0:00.10 sw-engine
8521 psaadm 20 0 371476 30712 19544 R 41.2 0.8 0:00.10 sw-engine

 

[vulcain]

I don't even need a site, everytime I access the panel for something the load goes to 20 and I need to reboot. going back to cpanel, it's beautiful but not working at all.

 

[TomBoB]

Hi,

KiB Mem : 3881672 total, 104716 free,

shows your server has only 4GB memory, which is barely more than the required minimum for Plesk. Hence me asking the above questions.
With only ~100MB memory free, it's no wonder it's not performing.

Have you tried
Plesk sw-engine process causes high CPU usage
to pin down what exactly is causing the high cpu load?

Cheers,Tom

 

[vulcain]

yes but was working fine before. has to do with the fpm setup

[root@server ~]# ps aux | grep sw-engine
psaadm 2354 10.8 4.7 636064 184640 ? S 00:58 0:08 sw-engine-fpm: pool plesk
psaadm 3438 1.4 1.3 521300 50752 ? S 00:59 0:00 sw-engine-fpm: pool plesk
psaadm 3440 10.8 4.3 635800 166964 ? S 00:59 0:02 sw-engine-fpm: pool plesk
root 3471 0.0 0.0 112664 972 pts/0 S+ 01:00 0:00 grep --color=auto sw-engine

 

[TomBoB]

Hi again,

did you turn on debugging and logging of sw-engine in /etc/sw-engine/sw-engine-fpm.conf ?
Any indications in the created log files?

Cheers, Tom

 

[vulcain]

it looks like the static mode is causing this issue, the ondemand is causing another one.
[15-Dec-2017 15:46:01.353826] DEBUG: pid 889, fpm_pctl_perform_idle_server_maintenance(), line 363: [pool plesk] currently 0 active children, 3 spare children
[15-Dec-2017 15:46:02.355159] DEBUG: pid 889, fpm_pctl_perform_idle_server_maintenance(), line 363: [pool plesk] currently 0 active children, 3 spare children
[15-Dec-2017 15:46:03.356473] DEBUG: pid 889, fpm_pctl_perform_idle_server_maintenance(), line 363: [pool plesk] currently 0 active children, 3 spare children
[15-Dec-2017 15:46:04.357885] DEBUG: pid 889, fpm_pctl_perform_idle_server_maintenance(), line 363: [pool plesk] currently 0 active children, 3 spare children
[15-Dec-2017 15:46:04.413505] DEBUG: pid 889, fpm_got_signal(), line 76: received SIGCHLD
[15-Dec-2017 15:46:04.413714] DEBUG: pid 889, fpm_children_bury(), line 254: [pool plesk] child 2250 has been killed by the process management after 87.183750 seconds from start
[15-Dec-2017 15:46:04.413758] DEBUG: pid 889, fpm_event_loop(), line 419: event module triggered 1 events
[15-Dec-2017 15:46:05.359448] DEBUG: pid 889, fpm_pctl_perform_idle_server_maintenance(), line 363: [pool plesk] currently 0 active children, 2 spare children
[15-Dec-2017 15:46:05.390518] DEBUG: pid 889, fpm_got_signal(), line 76: received SIGCHLD
[15-Dec-2017 15:46:05.390651] DEBUG: pid 889, fpm_children_bury(), line 254: [pool plesk] child 2339 has been killed by the process management after 68.083840 seconds from start
[15-Dec-2017 15:46:05.390679] DEBUG: pid 889, fpm_event_loop(), line 419: event module triggered 1 events
[15-Dec-2017 15:46:06.360878] DEBUG: pid 889, fpm_pctl_perform_idle_server_maintenance(), line 363: [pool plesk] currently 0 active children, 1 spare children
[15-Dec-2017 15:46:06.396442] DEBUG: pid 889, fpm_got_signal(), line 76: received SIGCHLD
[15-Dec-2017 15:46:06.396581] DEBUG: pid 889, fpm_children_bury(), line 254: [pool plesk] child 2340 has been killed by the process management after 69.072130 seconds from start
[15-Dec-2017 15:46:06.396609] DEBUG: pid 889, fpm_event_loop(), line 419: event module triggered 1 events
[15-Dec-2017 15:46:07.361197] DEBUG: pid 889, fpm_pctl_perform_idle_server_maintenance_heartbeat(), line 489: stopping idle maintenance as no children present ...

 

[danami]

I know that many providers just use the default install for Plesk which includes docker. I would check if docker is installed and if you aren't using it then I would uninstall it to free up resources.

 

[Bitpalast]

it looks like the static mode is causing this issue, the ondemand is causing another one....

The correct mode for servers with limited RAM is "ondemand". When you select "static" (of if static was brought in by an update), PHP instances are constantly kept in RAM, occupying a lot of RAM.

What "other" issue do you see when you change the setting to "ondemand"?