OOM Killer issue, container

[fropa]

Hi, I use Proxmox 5.1 with up to 70 containers. A few days ago one of the containers crashed. After some troubleshooting, I noticed that the Proxmox killed MariaDB service inside the container for memory reason.

Task in /lxc/170/ns killed as a result of limit of /lxc/170

memory: usage 677776kB, limit 1048576kB, failcnt 0

memory+swap: usage 1048576kB, limit 1048576kB, failcnt 15930052

kmem: usage 71520kB, limit 9007199254740988kB, failcnt 0

In monitoring systems and even on Proxmox web diagrams there is no such event. There I can see memory usage 600MB and the container has 400MB freer. So, what can I do to avoid that event in the future?

 

[fiona]

Hi,
while the container did not reach the memory limit, it did reach the memory+swap limit. Therefore it was killed. With cgroups v1 it is not possible to disable swapping in the container, except if you disable swapping on the host. You can try to monitor the processes running in the container for memory usage to determine the culprit.

 

[Paspao]

Hello,

I am having similar issues with our Proxmox 6 and LXC with Debian 10.

Sometimes I find stopped containers and sometimes only some daemons die, and I see lots of kernel messages in syslog.

Can someone suggest the fastest way to monitor processes for memory usage unattended , like logging until issue happens?

Is it better to monito processes on host or inside container?

Thank you.
P.