Guest machines regularly slow down

[renikrenik]

Good day!

Every day, between 12:30-13:00, guest cars start to slow down. Sites open slowly, it is difficult to work with documents, you make a payment in a bank for 5 minutes. It works slowly for 2-3 hours, and returns at 15-00.

What I tried:

check the load on the processor through "htop" and "top", did not see any anomalies (but there is little experience);

checked free memory, about 10-20 GB free;

checked the speed of the disk with the guest, more than 1000 Mb / s read / write;

checked disk and processor temperatures, no anomalies;

checked the load of the network channel, the load is not higher than 10 Mbps;

checked the channel load for users of guest machines, stable 15-20 Mbps;

checked the date and time on the guest, and the schedule of backups.


Help, please, how can I diagnose the problem?

 

[shanreich]

Can you check the pressure stats during the time this is happening?

head /proc/pressure/*

 

[renikrenik]

Today anomalies at 14-00:

root@proxmox1:~# head /proc/pressure/*
==> /proc/pressure/cpu <==
some avg10=10.56 avg60=26.88 avg300=13.88 total=3348138018
full avg10=0.61 avg60=0.22 avg300=0.33 total=2166910557

==> /proc/pressure/io <==
some avg10=0.00 avg60=0.09 avg300=0.15 total=479436109
full avg10=0.00 avg60=0.07 avg300=0.12 total=440090073

==> /proc/pressure/memory <==
some avg10=0.00 avg60=0.00 avg300=0.00 total=0
full avg10=0.00 avg60=0.00 avg300=0.00 total=0

root@proxmox1:~# head /proc/pressure/*
==> /proc/pressure/cpu <==
some avg10=2.29 avg60=19.55 avg300=13.03 total=3348426486
full avg10=0.94 avg60=0.44 avg300=0.38 total=2167149837

==> /proc/pressure/io <==
some avg10=0.00 avg60=0.06 avg300=0.14 total=479447082
full avg10=0.00 avg60=0.05 avg300=0.11 total=440099870

==> /proc/pressure/memory <==
some avg10=0.00 avg60=0.00 avg300=0.00 total=0
full avg10=0.00 avg60=0.00 avg300=0.00 total=0

 

[shanreich]

Seems like your load is getting too high during this timeframe:

==> /proc/pressure/cpu <==
some avg10=10.56 avg60=26.88 avg300=13.88 total=3348138018

Means that in the last 60 seconds 27% of processes were waiting for CPU time, which indicates that the CPU is under too high load.

Are you running any tasks using lots of CPU on the host / in any VM during that time?
Have you maybe overprovisioned your VMs wrt CPUs?

 

[renikrenik]

Thank you for the clarification!
But, I don't know what the guest machines do.

I can view graphs in the Summary tab, and I can view anything through the Shell.

I have 10 guest machines, I read the instructions, but did not find instructions on how to split the processor cores.

Please advise.
Now I will allocate 4 CPUs per guest.

My processor in the server:
CPU(s) 20 x Intel(R) Xeon(R) CPU E5-2666 v3 @ 2.90GHz (1 Socket)