| *** auk has joined #kata-general | 00:59 | |
| *** auk has quit IRC | 04:46 | |
| *** manny has quit IRC | 05:48 | |
| *** manny has joined #kata-general | 05:54 | |
| *** auk has joined #kata-general | 06:30 | |
| *** auk has quit IRC | 06:30 | |
| *** auk has joined #kata-general | 07:55 | |
| *** sameo has joined #kata-general | 08:10 | |
| *** sgarzare has joined #kata-general | 08:13 | |
| *** CeeMac has quit IRC | 08:22 | |
| *** gwhaley has joined #kata-general | 09:04 | |
| *** auk has quit IRC | 10:00 | |
| *** sameo has quit IRC | 10:54 | |
| *** lpetrut has joined #kata-general | 12:02 | |
| *** gwhaley has quit IRC | 12:24 | |
| *** gwhaley has joined #kata-general | 12:24 | |
| *** gwhaley has quit IRC | 12:41 | |
| *** gwhaley has joined #kata-general | 12:41 | |
| *** sameo has joined #kata-general | 13:30 | |
| *** sameo has quit IRC | 15:50 | |
| kata-irc-bot3 | <william> Is anyone here able to say something about the constraints of what applications can be ran using kata-firecracker? | 15:56 |
|---|---|---|
| kata-irc-bot3 | <graham.whaley> @william - hi - so, I hope so ;). And, yes, I'd quite like to see some of that captured on the kata github docs. /cc @eric.ernst @xu - can you help, or direct/nominate? | 16:03 |
| *** lpetrut has quit IRC | 16:13 | |
| kata-irc-bot3 | <william> I'll give some context if it helps: So far I have successfully ran some smaller applications in very basic images to benchmark kata-qemu and kata-fc. Now I am moving on towards application level benchmarks. When attempting to deploy https://github.com/DescartesResearch/TeaStore all pods successfully reach a running state but some applications will not progress and as I am currently debugging that I thought I'd ask the question | 16:13 |
| kata-irc-bot3 | above. When deleting my deployments my kata-fc pods also remains in termination mode indefinitely while this is not the case for kata-qemu. | 16:13 |
| kata-irc-bot3 | <william> @graham.whaley while I have you on the line. In reference to https://github.com/kata-containers/runtime/issues/2161 I did as suggested and added the kernel_params = "init=/usr/bin/kata-agent" to the kata config because my pods were running out of working memory. I am intrigued to know how that solved the issue and if it imposes any challenges. | 16:21 |
| kata-irc-bot3 | <graham.whaley> @william - ah, it was @julio.montes who suggested the use of the `init` fix - I'll let him wade in on that. /me goes to check Issue... ah, yes, I noted that the kernel `file max` value is set at boot according to how much RAM is in the system - so, this will be different between kata and a soft container (runc) - `runc` will effectively 'see' the whole of the host RAM, as it will be using/seeing the host kernel, whereas the | 16:30 |
| kata-irc-bot3 | kata vm kernel will only see the RAM passed to it, calculated according to the container/pod restrictions/requirements - and thus will likely always get a smaller `file max`, that reflects the actual RAM available to the pod/container. One could argue that that is actually more 'accurate' :slightly_smiling_face: | 16:30 |
| kata-irc-bot3 | <william> Right. To make the comparison as equal as possible I am using resource restrictions in my pod spec which to the best of my knowledge does a decent job of making sure each pod has the same resources no matter what RuntimeClass is chosen. Hopefully this will be a good indication of what each runtime is able to perform with the same amount of resources, then I'll just have to measure the resource overhead for the sandboxes and account | 16:38 |
| kata-irc-bot3 | for that :slightly_smiling_face: | 16:38 |
| *** sgarzare has quit IRC | 17:22 | |
| *** gwhaley has quit IRC | 18:09 | |
| *** lpetrut has joined #kata-general | 20:07 | |
| *** zer0def has quit IRC | 20:18 | |
| *** zer0def has joined #kata-general | 20:19 | |
| *** lpetrut has quit IRC | 21:03 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!