| *** auk has joined #kata-general | 00:48 | |
| *** auk has quit IRC | 02:02 | |
| *** lpetrut has joined #kata-general | 05:40 | |
| *** lpetrut has quit IRC | 07:04 | |
| *** sameo has joined #kata-general | 07:29 | |
| *** sgarzare has joined #kata-general | 07:37 | |
| *** sgarzare has quit IRC | 07:42 | |
| *** gwhaley has joined #kata-general | 08:01 | |
| kata-irc-bot2 | <graham.whaley> hi @kevin.fox @greg.bock - yeah, we have an 'incompatability' between Kata and minikube since Kata v1.6.2 - we switched networking model to use 'tc', and that does not seem to work under Minikube - see https://github.com/kata-containers/documentation/pull/445#issuecomment-489153009 | 08:21 |
|---|---|---|
| kata-irc-bot2 | <graham.whaley> I'm working with @archana.m.shinde to figure out exactly what under minikube is not working (still strongly suspect a kernel config setting), and then figure the best plan to 'fix it'. You can temporarily hand-fix the networking in kata back to `macvtap` in the kata config, and it should work - but, when you reboot your minikube `kata-deploy` will re-install with the default `tc` setup again. | 08:23 |
| *** sameo has quit IRC | 09:08 | |
| *** lpetrut has joined #kata-general | 09:30 | |
| *** sameo has joined #kata-general | 09:39 | |
| *** gwhaley has quit IRC | 11:00 | |
| *** gwhaley has joined #kata-general | 12:02 | |
| *** sgarzare has joined #kata-general | 13:24 | |
| *** lpetrut has quit IRC | 13:53 | |
| *** lpetrut has joined #kata-general | 14:15 | |
| *** lpetrut has quit IRC | 14:26 | |
| *** sameo has quit IRC | 14:49 | |
| kata-irc-bot2 | <kevin.fox> I can confirm that works. thanks. :slightly_smiling_face: | 15:49 |
| kata-irc-bot2 | <kevin.fox> A more specific error message might help debug that issue faster. | 15:49 |
| kata-irc-bot2 | <graham.whaley> @kevin.fox - the minikube issue that is? Yeah, it's hard though. afaik, the kernel just reports 'it did not work', with no details, and the actual 'index 4' error falls out of the go library that is doing the network diddling for us. Also, I don't think we've hit it before ;) | 15:51 |
| kata-irc-bot2 | <graham.whaley> Once we figure out what causes it, hopefully we can: | 15:51 |
| kata-irc-bot2 | <graham.whaley> - make the error more useful | 15:52 |
| kata-irc-bot2 | <kevin.fox> yeah. | 15:52 |
| kata-irc-bot2 | <graham.whaley> - auto-detect 'tc' support maybe, and adapt or error out nicely. | 15:52 |
| kata-irc-bot2 | <graham.whaley> If we can't fix this with the minikube kernel as-is then we may have to adapt kata-deploy to detect what the system supports and do the right thing | 15:52 |
| kata-irc-bot2 | <kevin.fox> autodetect tc would be cool. | 15:52 |
| kata-irc-bot2 | <kevin.fox> or just if it fails return 'network provisioning error' or something to point a user in the right direction to debug. | 15:53 |
| kata-irc-bot2 | <kevin.fox> I had no clue if it was virtualization, networking, storage, etc. | 15:53 |
| kata-irc-bot2 | <graham.whaley> well, we have to try and honour the ask in the config file as well - if somebody configs for 'tc' and it will not work, we'd probably not revert back out to 'macvtap' iyswim. Also, I would suggest an 'auto' mode where kata tries to pick the 'best supported setup', but then I expect in the real world people know exactly what their deployment stack looks like, and don't like surprises of things just 'working' due to auto setup, | 15:54 |
| kata-irc-bot2 | but maybe not working in the way they want, or obscuring deeper issues :slightly_smiling_face: | 15:54 |
| kata-irc-bot2 | <kevin.fox> yeah. thats why logging in that error case might be better in some ways. cause if they asked for tc and it failed, it should tell the user that tc specifically failed for some reason (or unknown) | 15:55 |
| kata-irc-bot2 | <graham.whaley> I'll /cc @archana.m.shinde on that ;) ^^ | 15:55 |
| kata-irc-bot2 | <kevin.fox> thanks. :slightly_smiling_face: | 15:55 |
| kata-irc-bot2 | <kevin.fox> I saw the emptyDir stuff in the change log for 1.7. thats why I was kicking the tires again. | 15:58 |
| kata-irc-bot2 | <kevin.fox> we've been considering adopting kata for a long time but haven't quite gotten the activation energy to deploy it. Fast docker in kata support would be the killer feature. | 15:59 |
| kata-irc-bot2 | <kevin.fox> we've had bad experience with devicemapper though so would rather avoid it. | 15:59 |
| kata-irc-bot2 | <kevin.fox> looks like the emptyDir thing isn't working like expected on 9p unfortunately. | 16:00 |
| kata-irc-bot2 | <eric.ernst> DinK (docker in kata) - is this for doing container image building, @kevin.fox? | 16:20 |
| kata-irc-bot2 | <archana.m.shinde> @kevin.fox What kind of issues have you run into with devicemapper? | 16:30 |
| kata-irc-bot2 | <kevin.fox> On a particular test machine which I was stressing it heavily, I ended up corrupting the pool of thin provisioned storage. | 16:45 |
| kata-irc-bot2 | <kevin.fox> now I'm gunshy of it. | 16:45 |
| kata-irc-bot2 | <kevin.fox> @eric.ernst yeah. safe container building is one of the few unanswered questions in k8s. | 16:45 |
| *** sameo has joined #kata-general | 16:52 | |
| *** gwhaley has quit IRC | 16:59 | |
| *** sgarzare has quit IRC | 17:08 | |
| *** lpetrut has joined #kata-general | 17:23 | |
| *** lpetrut has quit IRC | 17:27 | |
| kata-irc-bot2 | <raravena80> Dink heh.... how about KinK? (Kubernetes in Kata) Just like: https://github.com/kubernetes-sigs/kind | 17:45 |
| *** lcastell has joined #kata-general | 19:30 | |
| *** igordc has joined #kata-general | 21:37 | |
| *** igordc has quit IRC | 21:40 | |
| *** igordc has joined #kata-general | 21:44 | |
| *** sameo has quit IRC | 22:09 | |
| *** igordc has quit IRC | 23:55 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!