Closed WanzenBug closed 1 year ago
I'm fine with "backporting" that in general, but where does this come from? piraeus is not rhel/alma based, we used bullseye without issues, or did that pop up in bullseye now as well or where does this actually come from?
The issue pops up when running the container on an almalinux9 host. The file limit is not configured by the container OS but the host OS
dmeventd will be started by lvm commands (i.e. lvcreate, lvremove, ..) if it does not already run. This is an issue if started in a context with a high soft file limit: dmeventd will try to close all FDs after forking. The limit on Almalinux9 is higher than 1_000_000_000, which caused the initial command to wait for over 100 seconds, mostly leading to linstor commands timing out.
The workaround here is to start "dmeventd" before starting the satellite with a known reasonable file limit.
See also 0c6c6ec0d1d1d38320f66295d9f992ea8ee50815 in the linstor-server project.