*** LinuxMe has joined #kata-dev | 00:04 | |
*** LinuxMe has quit IRC | 00:09 | |
*** annabelleB has joined #kata-dev | 01:07 | |
*** annabelleB has quit IRC | 01:09 | |
*** fuentess has joined #kata-dev | 01:15 | |
*** lamego_ is now known as lamego | 01:16 | |
*** dlw has joined #kata-dev | 01:39 | |
*** LinuxMe_ has joined #kata-dev | 01:46 | |
*** annabelleB has joined #kata-dev | 02:11 | |
*** LinuxMe_ has quit IRC | 02:13 | |
*** annabelleB has quit IRC | 02:44 | |
*** lamego has left #kata-dev | 02:50 | |
*** LinuxMe has joined #kata-dev | 02:52 | |
*** LinuxMe has quit IRC | 02:57 | |
*** fuentess has quit IRC | 03:47 | |
*** LinuxMe has joined #kata-dev | 04:00 | |
*** LinuxMe has quit IRC | 04:05 | |
*** sjas has joined #kata-dev | 04:21 | |
*** sjas_ has quit IRC | 04:24 | |
*** LinuxMe has joined #kata-dev | 05:00 | |
*** LinuxMe has quit IRC | 05:04 | |
*** annabelleB has joined #kata-dev | 05:18 | |
*** annabelleB has quit IRC | 05:33 | |
*** LinuxMe has joined #kata-dev | 06:29 | |
*** LinuxMe has quit IRC | 06:35 | |
*** sameo has joined #kata-dev | 07:00 | |
*** LinuxMe has joined #kata-dev | 07:06 | |
*** LinuxMe has quit IRC | 07:11 | |
*** jodh has joined #kata-dev | 07:17 | |
*** davidgiluk has joined #kata-dev | 08:02 | |
*** gwhaley has joined #kata-dev | 08:05 | |
*** cdent has joined #kata-dev | 08:53 | |
*** LinuxMe has joined #kata-dev | 08:57 | |
*** LinuxMe has quit IRC | 09:02 | |
*** LinuxMe has joined #kata-dev | 09:34 | |
stefanha | Is Kata supposed to work with modular kernels? The rootfs Fedora script seems to be missing the systemd-udev package that will load drivers on demand. | 09:38 |
---|---|---|
*** LinuxMe has quit IRC | 09:39 | |
stefanha | lsm5: ^ | 09:43 |
stefanha | lsm5: Have you tweaked to the Fedora rootfs script? | 09:44 |
gwhaley | @stefanha - kernels, as in the VM kernels. Historically, ClearContainers did not support modules, but runv did iirc. I don't think we've enabled modules for Kata yet - but I think the agreement is that we probably will | 09:49 |
gwhaley | we would then I believe have to have a method (a mount) to get the modules dir mapped from the host into the guest, and ensure | 09:50 |
gwhaley | we have the right modules (versions) for that VM kernel. | 09:50 |
gwhaley | I have in the past suggested we should turn on module signing (the kernel module version check config), to try and ensure we don't try to load the wrong modules, which could cause a debug nightmare :-) | 09:50 |
stefanha | gwhaley: I see. The rootfs.sh KERNEL_MODULES_DIR= environment variable is working fine for me. I used it to copy in the necessary kernel modules. | 09:51 |
stefanha | kata-agent is running happily inside the VM with Fedora initramfs but for some reason docker + kata-runtime are waiting for something. Eventually creating the container times out. | 09:52 |
* stefanha looks at the runtime | 09:53 | |
* gwhaley not up to date on this - ah, so you can set that KERNEL_MODULES_DIR when doing an osbuilder rootfs build, and it will copy those modules into the rootfs image (rather than have them in a host side share that is mapped/mounted into the VM). That sounds like a good interim for allowing customised images with modules then. | 09:54 | |
stefanha | gwhaley: I've realized that the Fedora rootfs was not designed for initramfs. It's several hundred MB large :-). But at the moment this seems to work. | 09:55 |
stefanha | The last log messages from kata-runtime are "VM started" and "proxy started". Then the proxy logs kernel boot messages, followed by kata-agent startup messages. | 09:55 |
stefanha | After that, silence. | 09:55 |
* gwhaley goes to find ref for magic debug tricks... | 09:56 | |
stefanha | I have enabled them according to the Developer Guide. | 09:56 |
stefanha | Hopefully a backtrace from kata-runtime will contain useful info. | 09:56 |
gwhaley | then there is a debug console : https://github.com/kata-containers/documentation/blob/master/Developer-Guide.md#set-up-a-debug-console | 09:57 |
* gwhaley has not done that himself though | 09:57 | |
jodh | stefanha: yes, you can "kill -USR1 $runtime_pid" to get the backtrace but make sure you enable full debug for the system (see dev guide). But, if you're using the debug console, read that section carefully as you have to *disable* proxy debug for it to work. | 09:59 |
stefanha | jodh: Thanks! | 10:00 |
*** zerocoolback has joined #kata-dev | 10:09 | |
*** LinuxMe has joined #kata-dev | 10:11 | |
*** LinuxMe has quit IRC | 10:15 | |
stefanha | Is there a way to ping the agent from the host? | 10:29 |
stefanha | I see an HTTP2 request being sent by the proxy to kata.sock, | 10:30 |
stefanha | but no response. | 10:30 |
stefanha | (The agent has started and is running inside the guest.) | 10:30 |
gwhaley | @stfanha - not that I know of. /cc @jodh? | 10:31 |
*** zerocoolback has quit IRC | 10:31 | |
stefanha | It would be funny if Fedora's SELinux is stopping the agent :) | 10:33 |
stefanha | Will check that out | 10:33 |
*** zerocoolback has joined #kata-dev | 10:34 | |
jodh | stefanha: gwhaley: there is an api, but it's unused atm afaik - https://github.com/kata-containers/agent/blob/master/protocols/grpc/health.proto#L36 | 10:34 |
gwhaley | hey @jodh, whilst you are here ;-) - do you know if the '--log=filename' option works for the kata runtime (I presume it does) - I'm trying to enable it via /etc/docker/daemon.json adding a runtime args, but not seeing any output yet | 10:36 |
gwhaley | will try harder... | 10:36 |
jodh | gwhaley: are you trying to create separate kata-runtime logs for each "docker run"? | 10:37 |
gwhaley | not yet... just dumping them into a single file for a start. But, if it is easy to get separate ones for each run - then that would be great.... | 10:38 |
gwhaley | trying to fathom who is dropping log entries and where - trying to see if there is a difference between if we log to a file and what ends up in journald | 10:39 |
jodh | gwhaley: the --log= should work for the runtime, yes. The complication is passing that to the runtime with docker. The easiest method is to configure docker to call kata-runtime.sh and then do the --log= magic in there. | 10:39 |
gwhaley | right - I did consider using the 'via a shell script' trick ;-) I may have to go that route. Setting it in the docker daemon.json doesn't seem to be working for me | 10:40 |
jodh | gwhaley: ... but, no, you can't currently create separate logs I don't think. We could do this, but it's a bit gross as I think we'd need to look at a magic envvar in config.json. | 10:40 |
gwhaley | OK, I'll go shell route. The docker docs say there is a 'runtimeArgs' json var you can set to pass args to a specific runtime, but I'm not seeing a log file appear. np | 10:41 |
*** djinni has quit IRC | 10:42 | |
*** zerocoolback has quit IRC | 10:44 | |
*** dlw has quit IRC | 10:44 | |
*** LinuxMe has joined #kata-dev | 10:44 | |
*** zerocoolback has joined #kata-dev | 10:46 | |
*** LinuxMe has quit IRC | 10:48 | |
*** jugs has quit IRC | 10:51 | |
*** jugs has joined #kata-dev | 10:53 | |
*** djinni has joined #kata-dev | 10:54 | |
*** zerocoolback has quit IRC | 11:01 | |
*** zerocoolback has joined #kata-dev | 11:06 | |
*** gwhaley has quit IRC | 11:07 | |
*** LinuxMe has joined #kata-dev | 11:09 | |
*** LinuxMe_ has joined #kata-dev | 11:10 | |
*** LinuxMe has quit IRC | 11:14 | |
*** zerocoolback has quit IRC | 11:15 | |
*** LinuxMe_ has quit IRC | 11:15 | |
*** zerocoolback has joined #kata-dev | 11:16 | |
stefanha | For the record, the issue with a Fedora guest was that agent mis-detects the VMware vsock driver instead of using the virtio-serial port. | 11:18 |
stefanha | I or lsm5 will need to revisit this issue for Fedora. | 11:18 |
*** zerocoolback has quit IRC | 11:47 | |
*** zerocoolback has joined #kata-dev | 11:49 | |
*** devimc has joined #kata-dev | 11:51 | |
*** LinuxMe has joined #kata-dev | 12:02 | |
*** LinuxMe has quit IRC | 12:06 | |
*** zerocoolback has quit IRC | 12:18 | |
*** zerocoolback has joined #kata-dev | 12:18 | |
lsm5 | stefanha: no changes to rootfs yet from my side, atm working on setting up a COPR repo with my spec file changes | 12:29 |
*** gwhaley has joined #kata-dev | 12:35 | |
*** LinuxMe has joined #kata-dev | 12:38 | |
*** LinuxMe has quit IRC | 12:43 | |
*** fuentess has joined #kata-dev | 12:52 | |
*** LinuxMe has joined #kata-dev | 13:02 | |
stefanha | lsm5: FWIW I just got Kata running on Fedora using a Fedora 28 rootfs. | 13:03 |
*** LinuxMe has quit IRC | 13:03 | |
stefanha | It's a generic, modular kernel, not a ClearContainers kernel. | 13:03 |
stefanha | lsm5: I'll send you an email with details on the various things I needed to tweak. | 13:03 |
stefanha | lsm5: I'm just playing around with Kata, not trying to create RPM packages, in case you're wondering. | 13:05 |
stefanha | lsm5: But I guess you have or will hit the same issues as me when running it,. | 13:06 |
lsm5 | stefanha: ack, thanks, i have currently setup https://copr.fedorainfracloud.org/coprs/lsm5/kata/builds/ ..but only managed to get qemu-lite to build in copr, weird issues with other packages though they build just fine on fedora koji | 13:12 |
lsm5 | retrying with srpms instead of git sources | 13:14 |
*** LinuxMe has joined #kata-dev | 13:16 | |
*** LinuxMe has quit IRC | 13:17 | |
*** LinuxMe has joined #kata-dev | 13:17 | |
*** LinuxMe has quit IRC | 13:27 | |
*** LinuxMe has joined #kata-dev | 13:31 | |
*** lamego has joined #kata-dev | 13:31 | |
*** annabelleB has joined #kata-dev | 14:41 | |
*** zerocoolback has quit IRC | 14:54 | |
*** sameo has quit IRC | 15:06 | |
*** openstack has quit IRC | 15:22 | |
*** openstack has joined #kata-dev | 15:22 | |
*** ChanServ sets mode: +o openstack | 15:22 | |
*** zerocoolback has joined #kata-dev | 15:36 | |
*** zerocoolback has quit IRC | 15:37 | |
*** zerocoolback has joined #kata-dev | 15:38 | |
*** fiddletwix has quit IRC | 15:42 | |
*** zerocoolback has quit IRC | 15:42 | |
*** libregeekingkid[ has quit IRC | 16:09 | |
*** justJanne has quit IRC | 16:09 | |
*** mugsie has quit IRC | 16:09 | |
*** fiddletwix has joined #kata-dev | 16:13 | |
*** libregeekingkid[ has joined #kata-dev | 16:14 | |
*** justJanne has joined #kata-dev | 16:14 | |
*** mugsie has joined #kata-dev | 16:14 | |
clarkb | I had to explicitly insmod vhost_net on fedora 28 kernel as well | 16:14 |
clarkb | this wasn't necessary on ubuntu xenial | 16:14 |
*** libregeekingkid[ has quit IRC | 16:17 | |
*** annabelleB has quit IRC | 16:44 | |
*** libregeekingkid[ has joined #kata-dev | 17:00 | |
*** jodh has quit IRC | 17:11 | |
*** openstack has quit IRC | 17:11 | |
*** openstack has joined #kata-dev | 17:12 | |
*** ChanServ sets mode: +o openstack | 17:12 | |
*** LinuxMe has quit IRC | 17:14 | |
*** gwhaley has quit IRC | 17:30 | |
*** LinuxMe_ has joined #kata-dev | 17:37 | |
*** zerocoolback has joined #kata-dev | 17:40 | |
*** david-lyle has joined #kata-dev | 17:53 | |
*** dklyle has quit IRC | 17:56 | |
*** dklyle has joined #kata-dev | 17:57 | |
*** david-lyle has quit IRC | 17:58 | |
*** annabelleB has joined #kata-dev | 17:59 | |
*** zerocoolback has quit IRC | 18:28 | |
*** zerocoolback has joined #kata-dev | 18:28 | |
*** zerocoolback has quit IRC | 18:33 | |
*** annabelleB has quit IRC | 19:01 | |
*** annabelleB has joined #kata-dev | 19:04 | |
*** annabelleB has quit IRC | 19:07 | |
*** davidgiluk has quit IRC | 19:26 | |
*** annabelleB has joined #kata-dev | 19:36 | |
*** annabelleB has quit IRC | 19:41 | |
*** LinuxMe_ has quit IRC | 19:44 | |
*** cdent has quit IRC | 20:09 | |
clarkb | salvador.fuentes: as an updated the node failure issue is related to ipv6 networking in vexxhost. mnaser is going to look at it | 20:35 |
clarkb | *as an update | 20:36 |
*** devimc has quit IRC | 21:27 | |
*** fuentess has quit IRC | 21:35 | |
*** LinuxMe has joined #kata-dev | 21:38 | |
*** LinuxMe has quit IRC | 21:42 | |
*** lamego has left #kata-dev | 21:45 | |
*** LinuxMe has joined #kata-dev | 21:56 | |
*** fuentess has joined #kata-dev | 22:14 | |
*** LinuxMe has quit IRC | 22:29 | |
*** fuentess has quit IRC | 23:03 | |
*** devimc has joined #kata-dev | 23:31 | |
*** LinuxMe has joined #kata-dev | 23:36 | |
*** LinuxMe has quit IRC | 23:41 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!