repeating vmstat trap divide errors

[hdkarl]

Im using Proxmox VE 6.1-7 on Debian 10, all updates installed.

My /var/log/kern.log is full of trap divide errors, thrown every 5 minutes:
Feb 17 22:12:42 pmx kernel: [412584.239291] traps: vmstat[8166] trap divide error ip:401c8c sp:7ffdeff9b140 error:0 in vmstat[400000+5000]
Feb 17 22:17:34 pmx kernel: [412876.104863] traps: vmstat[16736] trap divide error ip:804b35a sp:ffcd92a0 error:0 in vmstat[8048000+5000]
Feb 17 22:17:53 pmx kernel: [412894.540959] traps: vmstat[17227] trap divide error ip:401c8c sp:7ffcd4f35b10 error:0 in vmstat[400000+5000]
Feb 17 22:22:53 pmx kernel: [413194.830005] traps: vmstat[26580] trap divide error ip:401c8c sp:7ffd931a4530 error:0 in vmstat[400000+5000]
Feb 17 22:27:55 pmx kernel: [413496.683343] traps: vmstat[2738] trap divide error ip:804b35a sp:ff883540 error:0 in vmstat[8048000+5000]
Feb 17 22:28:03 pmx kernel: [413505.118014] traps: vmstat[3051] trap divide error ip:401c8c sp:7ffdf42a8480 error:0 in vmstat[400000+5000]
Feb 17 22:32:55 pmx kernel: [413796.957186] traps: vmstat[11088] trap divide error ip:804b35a sp:ff8e03a0 error:0 in vmstat[8048000+5000]
Feb 17 22:33:14 pmx kernel: [413815.412612] traps: vmstat[11621] trap divide error ip:401c8c sp:7ffe1d07e9f0 error:0 in vmstat[400000+5000]

Is this some known bug? Is a solution available?

BR

 

[Stoiko Ivanov]

This is odd - and seems related to a division by 0 problem in (some version) of `vmstat`
* do you get such a message in the journal/kern.log if you run `vmstat` in a shell (if yes please post the output of `which vmstat` )?
* do you have any cronjob or monitoring running, which would run things once every 5 minutes?
** if yes what is running in those jobs?

I hope this helps!

 

[hdkarl]

I'm using /usr/bin/vmstat from procps 2:3.3.15-2 amd64

- running "vmstat 1" from a shell: no errors over more than 30 minutes
- cronjobs / monitoring: no matches for the timing pattern

 

[Stoiko Ivanov]

maybe anything running inside of containers is calling vmstat?

 

[hdkarl]

Yes! I found it.
The errors were triggered from 2 old Debian 6 containers, running webmin !

When "Collect system status in background" is enabled (by default) then webmin calls vmstat each 5 minutes.

And - I have no idea why - on two containers vmstat triggers a SIGFPE, using strace that happens after reading from /proc/vmstat and before writing its output.

 

[Stoiko Ivanov]

hmm - in my experience this can happen if your hypervisor is quite ahead of your lxc-container (or the other way around) - the libc needs to be somewhat in sync with the kernel interfaces.

consider upgrading the containers to a still supported version of debian (for security concerns)
If you really need to run Debian 6 a VM guest is probably a better option

I hope this helps!