| kata-irc-bot | <chelsea.e.mafrica> I ran into this same problem modifying a marginally related test. I created an issue: https://github.com/kata-containers/kata-containers/issues/4164 | 00:03 |
|---|---|---|
| kata-irc-bot | <cameron4mcdermott> Hi all, I have just opened an issue in the Github (https://github.com/kata-containers/kata-containers/issues/4167) related to clock drift. We have noticed that Kata pods running for a number of days seem to be slowly moving out of sync with NTP time, whereas new pods in Kata start with the correct time. Some debugging showed that the chronyd.service was failing on boot inside the guest due to `ReadWritePaths=/run | 12:01 |
| kata-irc-bot | /var/lib/chrony -/var/log`, but changing the middle path from `/var/lib/chrony` to `/var/run` was enough to get the service to start. I have yet to confirm if this fixes the drift but some feedback on the situation would be appreciated! | 12:01 |
| kata-irc-bot | <fidencio> Would be good to understand what's the guest image being used (if possible, please, add it to the issue) | 12:11 |
| kata-irc-bot | <cameron4mcdermott> The guest image is the default one shipped in the kata-static release tar. No modifications on my end apart from mounting it manually after install to a loopback device in order to copy across strace and alter the chronyd.service to find a working solution. Ill add this to the issue now | 13:01 |
| kata-irc-bot | <fidencio> Are you familiar with the kata-containers codebase? | 13:06 |
| kata-irc-bot | <cameron4mcdermott> To an extent yes | 13:07 |
| kata-irc-bot | <fidencio> We already do some changes in the chrony service directly, as you can see in tools/osbuilder/rootfs-builder/rootfs.sh | 13:07 |
| kata-irc-bot | <fidencio> If the distro requires changes, the way to go would be opening an issue for the distro themselves, but that could take a while to be brought back down. So, one approach would be to, maybe, edit the service as part of that script | 13:08 |
| kata-irc-bot | <cameron4mcdermott> okay this sounds like a suitable approach, I can extend the sed thats already been performed on the service to include the changes above as well | 13:11 |
| kata-irc-bot | <fidencio> Exactly. Would you be comfortable enough to submit a patch with that? | 13:16 |
| kata-irc-bot | <cameron4mcdermott> yes I am more than happy to do that | 13:20 |
| kata-irc-bot | <fidencio> Thanks, very much appreciated! <3 | 13:21 |
| kata-irc-bot | <andrew.czopiw> Hi everyone. Rook Ceph is a storage orchestration project for K8s. Does anyone know if Kata Containers is compatible with the Rook Ceph project and Multus networking ? Has anyone successfully used Kata Containers with Rook Ceph and Multus ? | 15:08 |
| kata-irc-bot | <andrew.czopiw> Hi everyone. Rook Ceph is a storage orchestration project for K8s. Does anyone know if Kata Containers is compatible with the Rook Ceph project and Multus networking ? Has anyone successfully used Kata Containers with Rook Ceph and Multus ? | 15:11 |
| kata-irc-bot | <cameron4mcdermott> https://github.com/kata-containers/kata-containers/pull/4168 pr created, let me know how it is when you get the chance! | 16:59 |
| kata-irc-bot | <fidencio> @cameron4mcdermott, fixing the indentation in the patch is needed, and after that I think we'd be good to go after that, as we got an okay from a clear linux dev. Also, it seems that the DCO check is failing because of a difference in the author (Champ-Goblem) and the signed-off-by person (Cameron McDermott). Last but not least, `Fixes: #4167` is the way to make the other failing check pass (instead of the full URL) | 18:22 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!