Question Health Monitor CPU usage "out of scale"

[Andrea Bampi]

I've experienced a very weird issue on two different servers (with different Linux distributions and Plesk versions) the same day (3 hours difference).
The most worrying occurred on a Ubuntu 14.04.5 LTS with Plesk 12.0.18: the Health Monitor graph shows a crazy CPU peak (see attachment) where the full-scale value should be 100 and it got up to 9.0M (!).
Just after the peak, the server shut down and rebooted - no issue whatsoever since then.
I checked all logs and didn''t find anything unusual (I surely may have missed something!), so I suspect a hardware/power issue.
Any clues of what may have happened? It never happened anything remotely similar (the server is up since november 2016...).
...the other server had a similar peak (900K on the graph) but didn't shut down. Nothing enlightening found in the logs...

 

[Bitpalast]

Is this a dedicated server or a virtual server?

 

[Andrea Bampi]

Is this a dedicated server or a virtual server?

Both dedicated physical servers on Register.it (hosting about 20 websites each).

 

[Bitpalast]

A sudden reboot could maybe be caused by a Linux kernel issue. The "9.0 M" is an unrealistic symptom. It only indicates that the CPU was unable to work on any more task when the problem occured, it is not really a measurement.

My suggestion is to first update Plesk to Plesk Onyx 17.5. This will enable you to use a newer version of your OS. Then update the Linux kernel to the latest version and update the rest of your OS to the latest version, too. Please see this page for details on the compatibility: Software Requirements for Plesk

Besides that, did you ask your data center whether known hardware infrastructure issues (with the power supply) are known for the times when the issues occured?

 

[Andrea Bampi]

Thanks for the suggestions. Unfortunately updating Plesk and especially OS is out of question (for now): several web applications and critical sites are involved. Besides, the server worked flawlessly for more than 1 year (and still is, since the "issue"), that's why I'm particularly puzzled.
The very first thing I did was opening a ticket on the support channel of my hoster: the datacenter guys don't SEEM to know anything about infrastructural issues. They proposed to investigate further by shutting down the server for a couple of hours to check the hardware. That's why I came here - I have to decide how to proceed...