| *** tmhoang has quit IRC | 06:45 | |
| *** igordc has quit IRC | 06:47 | |
| *** tmhoang has joined #kata-dev | 07:41 | |
| *** sgarzare has joined #kata-dev | 07:57 | |
| *** jodh has joined #kata-dev | 07:58 | |
| *** tmhoang has quit IRC | 08:15 | |
| *** sameo_ has joined #kata-dev | 08:27 | |
| *** tmhoang has joined #kata-dev | 08:45 | |
| *** davidgiluk has joined #kata-dev | 08:58 | |
| *** gwhaley has joined #kata-dev | 09:03 | |
| *** sameo_ has quit IRC | 10:15 | |
| *** sameo has joined #kata-dev | 10:15 | |
| *** sgarzare has quit IRC | 10:34 | |
| *** sgarzare has joined #kata-dev | 10:37 | |
| *** eguan has joined #kata-dev | 11:38 | |
| eguan | Hi, does kata-runtime support kubernetes postStart/preStop HTTP hook? I verified exec hook works as expected, but I didn't see http hook being triggered. Is this a known issue of kata-runtime? | 11:41 |
|---|---|---|
| kata-irc-bot | <xu> I guess there will be a networking issue | 11:43 |
| eguan | I created two pods, one is http server, another test pod has postStart http hook defined to POST to the http server, and the POST didn't trigger. But I can exec into the test pod and use curl to connect to the server without problems | 11:46 |
| eguan | So postStart/preStop HTTP hook is supposed to work with kata, right? | 11:49 |
| *** gwhaley has quit IRC | 11:57 | |
| *** tmhoang has quit IRC | 13:06 | |
| *** devimc has joined #kata-dev | 13:07 | |
| *** gwhaley has joined #kata-dev | 13:19 | |
| *** tmhoang has joined #kata-dev | 13:30 | |
| *** devimc has quit IRC | 13:55 | |
| *** fuentess has joined #kata-dev | 13:57 | |
| *** fuentess has quit IRC | 13:58 | |
| *** devimc has joined #kata-dev | 13:59 | |
| *** fuentess has joined #kata-dev | 14:00 | |
| kata-irc-bot | <graham.whaley> Hi @salvador.fuentes - just fyi, I've just moved the metrics CI slave from a t1 to a c1 instance on jenkins - I may have to do a few tweaks to get it running yet, but should be active soon. Let's hope that settles the metrics down.... | 15:18 |
| kata-irc-bot | <salvador.fuentes> Hi @graham.whaley, cool, thanks. Is that c1 instance a bigger machine? | 15:24 |
| kata-irc-bot | <graham.whaley> it is :slightly_smiling_face: but, we were not really after bigger - really wanted more stable. So, it is still 4 cpu, but a xeon with 32GB RAM. stability in my initial testing looks much better though - much more repeatable | 15:25 |
| kata-irc-bot | <salvador.fuentes> ahh ok, good, thanks for the explanation | 15:26 |
| *** devimc_ has joined #kata-dev | 16:34 | |
| *** devimc has quit IRC | 16:36 | |
| kata-irc-bot | <sebastien.boeuf> davidgiluk: thanks for pulling Maxime into the hotplug mem + vhost-user discussion as I think it'd be nice to have a complete and clear understanding about this use case. Kata is highly concerned by those potential issues | 17:02 |
| davidgiluk | sebastien: Nod; just checking but it's only really hot-plug you're interested in, not unplug? | 17:03 |
| kata-irc-bot | <graham.whaley> afaik, we don't enable hotunplug, coz it's hard and expensive :slightly_smiling_face: but, will let @sebastien.boeuf and maybe @jose.carlos.venegas.m confirm/deny that | 17:10 |
| davidgiluk | nod, it's the 'hard' bit that I was thinking of | 17:11 |
| kata-irc-bot | <sebastien.boeuf> davidgiluk: yeah from a Kata perspective, we're never performing unplug, only using ballooning for this. But in case of hotplug, I have tested and we run into an issue with vhost user, because if you try to read a file after the new memory has been added, the kernel might pick some of the new addresses that are unknown from the vhost-user backend, to allocate some memory pointed by the virtqueues | 17:13 |
| davidgiluk | yeh | 17:14 |
| kata-irc-bot | <sebastien.boeuf> davidgiluk: But I really want to understand if that's something fixable (meaning not implemented by QEMU, but feasible), or if there is a hardstop from the vhost-user protocol itself | 17:14 |
| kata-irc-bot | <sebastien.boeuf> it looks like we should be able to support this, but if Maxime can confirm, that'd be great | 17:14 |
| davidgiluk | sebastien: Yeh, I need to see what Maxime says; I think the idea is that what you do is you receive a new set-mem-table and go 'oh, it's all the same except for ---> that <---- mapping' and only change that one | 17:15 |
| kata-irc-bot | <eric.ernst> @james.o.hunt @graham.whaley et al, curious about your thoughts on: https://github.com/kata-containers/packaging/pull/371 | 17:20 |
| kata-irc-bot | <eric.ernst> probot's stale application in GitHub would be helpful, imo. | 17:20 |
| kata-irc-bot | <eric.ernst> threw this PR for packaging repo to get feedback. | 17:20 |
| kata-irc-bot | <eric.ernst> sinc most of the items are stale from me, i figured its a good starting place to try it. | 17:21 |
| kata-irc-bot | <eric.ernst> :slightly_smiling_face: | 17:21 |
| kata-irc-bot | <graham.whaley> hi @eric.ernst - was on my radar for today - let me go have a look. I'm guessing you using the probot hosting for that (not an action, or just testing with a beta action?). That would probably be my first question - who/how is it being hosted, and can we rely on that to stay around (think pullapprove etc.) | 17:21 |
| kata-irc-bot | <eric.ernst> GitHub hosted. | 17:22 |
| kata-irc-bot | <eric.ernst> @krsna1729 ^ | 17:22 |
| kata-irc-bot | <krsna1729> their own hosting. we can probably fork and point github actions at it | 17:23 |
| kata-irc-bot | <krsna1729> to be under our control | 17:23 |
| *** tmhoang has quit IRC | 17:24 | |
| kata-irc-bot | <krsna1729> this seemed like mild enough use of automation | 17:24 |
| kata-irc-bot | <krsna1729> for the hosting option to be less relevant since it does not touch code | 17:25 |
| kata-irc-bot | <eric.ernst> Yeah, it only has access to github issue metadata and being able to comment on issues. | 17:27 |
| kata-irc-bot | <eric.ernst> Seemed like good functionality with limited scope, imo | 17:27 |
| kata-irc-bot | <graham.whaley> @krsna1729 indeed, if it did go away nothing would break, things would just not auto-close, which is where we are at today. | 17:31 |
| *** sameo has quit IRC | 17:35 | |
| kata-irc-bot | <graham.whaley> @salvador.fuentes @sebastien.boeuf - the metrics CI is still not stable. When I ran by hand it was very stable, but now in the CI itself I think I'm seeing variations. I'll have to check again in the morning to see if they are expected or (probably not) being caused by the PRs that are being processed.... | 18:04 |
| kata-irc-bot | <sebastien.boeuf> :( | 18:06 |
| kata-irc-bot | <sebastien.boeuf> any idea what's causing those results to be not stable? | 18:06 |
| kata-irc-bot | <graham.whaley> @sebastien.boeuf not yet - when I ran 6 in a row yesterday on that same instance, they were rock solid. so, it could be something we do during the CI run that throws them out, or there is a small chance it just happens to be the PRs I tested with that throw the metrics out (doubtful). So, tomorrow I will gather the last 5-10 runs and look at their results and logs - if the results are not stable, then I'm going to have to think | 18:09 |
| kata-irc-bot | hard about what might be causing that... | 18:09 |
| *** jodh has quit IRC | 18:09 | |
| kata-irc-bot | <graham.whaley> it's when we get changing PSS footprint measures that I find confusing - unless something else is runnng and sharing a library or something, then I would have expected them to be fairly stable. Maybe I can add sanity checks to that code. | 18:10 |
| kata-irc-bot | <sebastien.boeuf> this sounds like a good idea | 18:11 |
| *** gwhaley has quit IRC | 18:12 | |
| *** sameo has joined #kata-dev | 18:43 | |
| kata-irc-bot | <eric.ernst> we should probably take another look @ https://github.com/kata-containers/runtime/issues/1203 now and see what's landed in distros now. | 18:44 |
| kata-irc-bot | <eric.ernst> anyone have a few minutes to check and verify we're okay on Ubuntu again? https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1813934a | 18:46 |
| openstack | Launchpad bug 1813934 in linux (Ubuntu Cosmic) "Vsock connect fails with ENODEV for large CID" [High,Fix committed] | 18:46 |
| kata-irc-bot | <eric.ernst> afaict its getting close but isn't released yet for bionic. | 18:52 |
| *** sgarzare has quit IRC | 19:12 | |
| kata-irc-bot | <jose.carlos.venegas.m> yes there is not unplug plans | 19:14 |
| kata-irc-bot | <jose.carlos.venegas.m> we have a pending PR that does balloon only to reduce memory, longer term is use virt-mem | 19:15 |
| kata-irc-bot | <sebastien.boeuf> @jose.carlos.venegas.m so virtio-mem readiness is not clear to me | 19:26 |
| kata-irc-bot | <sebastien.boeuf> that's why we should ask RH maintainers about this | 19:26 |
| kata-irc-bot | <sebastien.boeuf> and we should also invite them to talk during our AC meeting to present about this | 19:27 |
| kata-irc-bot | <sebastien.boeuf> this would be very nice, since I think the virtio-fs one from Stefan was very helpful to get everybody n the same page | 19:28 |
| kata-irc-bot | <jose.carlos.venegas.m> @sebastien.boeuf agree, in but agree probably to simplify the adoption we should go first feature by feature | 19:29 |
| kata-irc-bot | <archana.m.shinde> eguan Yes post start hooks are supposed to work. I verified this with running curl as a post-start hook. If you are still seeing issues, can you open a github issue with the pod yaml that you are using? | 19:29 |
| kata-irc-bot | <sebastien.boeuf> @jose.carlos.venegas.m yes agreed. For now the focus is virtio-fs, but you raised a good point with virtio-mem. Let me send an email to the maintainer to see what he thinks about presenting to us | 19:30 |
| kata-irc-bot | <jose.carlos.venegas.m> awesome! | 19:38 |
| kata-irc-bot | <sebastien.boeuf> @jose.carlos.venegas.m ok just sent it! if we can make this happen, we should make sure to spend more time on the topic before the presentation so that we can come up with some good questions | 19:45 |
| *** eernst has joined #kata-dev | 19:54 | |
| *** eernst has quit IRC | 19:54 | |
| *** davidgiluk has quit IRC | 20:17 | |
| *** sgarzare has joined #kata-dev | 20:20 | |
| *** sgarzare has quit IRC | 20:20 | |
| *** devimc_ has quit IRC | 20:23 | |
| *** devimc has joined #kata-dev | 20:25 | |
| *** sameo has quit IRC | 20:38 | |
| *** devimc has quit IRC | 20:44 | |
| *** tmhoang has joined #kata-dev | 20:56 | |
| *** devimc has joined #kata-dev | 21:16 | |
| *** eernst has joined #kata-dev | 22:06 | |
| *** eernst has quit IRC | 22:06 | |
| *** eernst has joined #kata-dev | 22:07 | |
| *** eernst has quit IRC | 22:25 | |
| *** fuentess has quit IRC | 22:25 | |
| *** devimc has quit IRC | 22:28 | |
| *** eernst has joined #kata-dev | 23:00 | |
| *** eernst has quit IRC | 23:02 | |
| *** eernst has joined #kata-dev | 23:06 | |
| *** eernst has quit IRC | 23:10 | |
| *** eernst has joined #kata-dev | 23:16 | |
| *** eernst has quit IRC | 23:17 | |
| *** eernst has joined #kata-dev | 23:17 | |
| *** eernst has quit IRC | 23:19 | |
| *** eernst has joined #kata-dev | 23:23 | |
| *** eernst has quit IRC | 23:27 | |
| *** tmhoang has quit IRC | 23:27 | |
| *** eernst has joined #kata-dev | 23:31 | |
| *** eernst has quit IRC | 23:32 | |
| *** dklyle has quit IRC | 23:39 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!