| kata-irc-bot | <anastassios.nanos> so, following up on this one -- it seems I was able to reproduce it locally: building all available rootfs images and testing with all of them with qemu & firecracker leads to the clearlinux agent_init=no case to fail only for firecracker. All other rootfs images (with the exception of debian where we're missing a package for the correct rootfs build) work fine, both with agent_init=yes/no | 06:26 |
|---|---|---|
| kata-irc-bot | <anastassios.nanos> so I'm thinking of adding our current static build rootfs image to the ci test process (instead of testing the ubuntu one -- I believe fabiano has opened a ticket already for that), see that it fails, and then we have a number of options -- dig into the reason this thing fails (why systemd segfaults in a nested firecracker VM in clearlinux), change our default rootfs image (try ubuntu/alpine), change agent_init=yes for the | 06:28 |
| kata-irc-bot | rootfs image etc. | 06:28 |
| kata-irc-bot | <fidencio> I think using systemd instead of the agent would be slower, not much, but at least one more process to be started (so then the agent could be started) | 06:30 |
| kata-irc-bot | <fidencio> https://github.com/kata-containers/tests/issues/4850 is the issue | 06:30 |
| kata-irc-bot | <fidencio> I think I'd be fine with testing using alpine, which is a way smaller rootfs than the ubuntu one, and then proposing the switch for that | 06:31 |
| kata-irc-bot | <jlineaweaver> Hello, I’m currently using Kata Containers and Firecracker, and I’m investigating retrieving metrics from them. I’ve found this link how-to-set-prometheus-in-k8s.md. However, the last change to it was a year ago. Just wondering if this is still the recommended way to monitor Kata in k8s? | 17:07 |
| kata-irc-bot | <jlineaweaver> Looking in quay.io (https://quay.io/organization/kata-containers) I don’t see the kata-monitor container | 17:15 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!