| *** mordred has quit IRC | 00:06 | |
| *** dhellmann has quit IRC | 00:06 | |
| *** dhellmann has joined #kata-dev | 00:11 | |
| *** ChanServ has quit IRC | 00:12 | |
| *** mordred has joined #kata-dev | 00:16 | |
| *** ChanServ has joined #kata-dev | 00:33 | |
| *** barjavel.freenode.net sets mode: +o ChanServ | 00:33 | |
| *** libregeekingkid[ has quit IRC | 00:38 | |
| *** JerryLiu has quit IRC | 00:44 | |
| *** JerryLiu has joined #kata-dev | 00:44 | |
| *** libregeekingkid[ has joined #kata-dev | 00:59 | |
| *** dlw has joined #kata-dev | 01:08 | |
| *** mordred has quit IRC | 01:31 | |
| *** mordred has joined #kata-dev | 01:45 | |
| *** dlw has quit IRC | 01:48 | |
| *** dlw has joined #kata-dev | 01:48 | |
| *** zerocoolback has joined #kata-dev | 02:26 | |
| *** dklyle has joined #kata-dev | 02:40 | |
| *** dlw has quit IRC | 02:44 | |
| *** dlw has joined #kata-dev | 02:44 | |
| *** fuentess1 has quit IRC | 02:48 | |
| *** dklyle has quit IRC | 02:54 | |
| *** zerocoolback has quit IRC | 03:13 | |
| *** zerocoolback has joined #kata-dev | 03:18 | |
| *** zerocoolback has joined #kata-dev | 03:18 | |
| *** zerocoolback has quit IRC | 03:20 | |
| *** zerocoolback has joined #kata-dev | 03:32 | |
| *** zerocoolback has quit IRC | 04:30 | |
| *** zerocoolback has joined #kata-dev | 05:01 | |
| *** zerocoolback has joined #kata-dev | 05:13 | |
| *** zerocoolback has quit IRC | 05:27 | |
| *** zerocoolback has joined #kata-dev | 05:28 | |
| *** zerocoolback has quit IRC | 05:32 | |
| *** zerocoolback has joined #kata-dev | 05:51 | |
| *** zerocoolback has quit IRC | 06:05 | |
| *** zerocoolback has joined #kata-dev | 06:06 | |
| *** zerocoolback has quit IRC | 06:06 | |
| *** zerocoolback has joined #kata-dev | 06:06 | |
| *** zerocoolback has quit IRC | 06:07 | |
| *** zerocoolback has joined #kata-dev | 06:07 | |
| *** zerocoolback has quit IRC | 06:07 | |
| *** zerocoolback has joined #kata-dev | 06:08 | |
| *** zerocoolback has quit IRC | 06:08 | |
| *** zerocoolback has joined #kata-dev | 06:09 | |
| *** zerocoolback has quit IRC | 06:09 | |
| *** jodh has joined #kata-dev | 07:31 | |
| *** jodh has quit IRC | 07:31 | |
| *** jodh has joined #kata-dev | 07:31 | |
| *** sameo has joined #kata-dev | 07:45 | |
| *** davidgiluk has joined #kata-dev | 07:59 | |
| *** gwhaley has joined #kata-dev | 08:01 | |
| *** dlw1 has joined #kata-dev | 09:08 | |
| *** dlw has quit IRC | 09:10 | |
| *** dlw1 is now known as dlw | 09:10 | |
| *** dlw has quit IRC | 09:42 | |
| *** dlw has joined #kata-dev | 09:58 | |
| * gwhaley wonders if the slack irc bot is working - I got an email notification from slack, but saw nothing here.... | 10:10 | |
| gwhaley | @gnawux @jodh - for help on setting up ARM CI - pls open an Issue on https://github.com/kata-containers/ci and copy in grahamwhaley and chavafg (fuentess) | 10:11 |
|---|---|---|
| gwhaley | you can see the Jenkins xml configs etc. in that repo, but unless you know how to read them, then that may or may not help (I don't know if copying them into a Jenkins setup and then tweaking them will work - maybe!) | 10:12 |
| *** dlw has quit IRC | 10:20 | |
| *** dlw has joined #kata-dev | 10:28 | |
| *** LinuxMe has joined #kata-dev | 10:31 | |
| *** LinuxMe has quit IRC | 10:35 | |
| *** sameo has quit IRC | 10:44 | |
| *** dlw has quit IRC | 10:52 | |
| *** dlw has joined #kata-dev | 10:54 | |
| *** dlw has quit IRC | 11:00 | |
| *** gwhaley has quit IRC | 11:06 | |
| *** mars704 has quit IRC | 11:40 | |
| *** zerocoolback has joined #kata-dev | 12:00 | |
| *** devimc has joined #kata-dev | 12:10 | |
| *** sameo has joined #kata-dev | 12:13 | |
| *** fuentess has joined #kata-dev | 12:29 | |
| *** gwhaley has joined #kata-dev | 12:34 | |
| *** LinuxMe has joined #kata-dev | 12:58 | |
| *** zerocoolback has quit IRC | 13:45 | |
| *** zerocool_ has joined #kata-dev | 13:47 | |
| *** sameo has quit IRC | 13:58 | |
| *** fiddletwix has quit IRC | 14:19 | |
| *** fiddletwix has joined #kata-dev | 14:20 | |
| *** zerocool_ has quit IRC | 14:22 | |
| *** dklyle has joined #kata-dev | 14:33 | |
| *** eernst has joined #kata-dev | 14:46 | |
| *** jodh has quit IRC | 14:49 | |
| *** jodh has joined #kata-dev | 14:49 | |
| *** jodh has quit IRC | 14:50 | |
| *** jodh has joined #kata-dev | 14:50 | |
| *** sameo has joined #kata-dev | 15:03 | |
| *** jodh has quit IRC | 15:04 | |
| *** jodh has joined #kata-dev | 15:04 | |
| *** jodh has quit IRC | 15:04 | |
| *** jodh has joined #kata-dev | 15:04 | |
| *** devimc has quit IRC | 15:05 | |
| *** devimc has joined #kata-dev | 15:23 | |
| *** LinuxMe has quit IRC | 16:07 | |
| *** LinuxMe has joined #kata-dev | 16:09 | |
| *** mcastelino has joined #kata-dev | 16:12 | |
| *** jodh has quit IRC | 16:57 | |
| *** gwhaley has quit IRC | 17:33 | |
| *** LinuxMe has quit IRC | 17:45 | |
| *** LinuxMe has joined #kata-dev | 17:46 | |
| *** kata-dev-irc-bot has joined #kata-dev | 18:31 | |
| kata-dev-irc-bot | <eric.ernst> @jose.carlos.venegas.m - hey carlos - I wanted to talk about release "stuff" | 18:56 |
| kata-dev-irc-bot | <eric.ernst> around? | 18:56 |
| *** JerryLiu has quit IRC | 19:09 | |
| *** JerryLiu has joined #kata-dev | 19:10 | |
| *** davidgiluk has quit IRC | 19:23 | |
| kata-dev-irc-bot | <niteshkonkar007> So, IIUC kata is moving its CI from Travis to Jenkins. | 19:31 |
| *** zerocoolback has joined #kata-dev | 19:32 | |
| kata-dev-irc-bot | <eric.ernst> @niteshkonkar007 no not really. | 19:39 |
| kata-dev-irc-bot | <eric.ernst> at least, I didn't think things were moving :slightly_smiling_face: | 19:39 |
| kata-dev-irc-bot | <eric.ernst> We use jenkins today. | 19:39 |
| kata-dev-irc-bot | <eric.ernst> @niteshkonkar007 - I always have a hard time tracking who's github matches who's name matches whos company/work, etc. | 19:40 |
| kata-dev-irc-bot | <eric.ernst> You @ IBM? | 19:40 |
| kata-dev-irc-bot | <niteshkonkar007> @eric.ernst: Yes | 19:40 |
| kata-dev-irc-bot | <eric.ernst> Cool. I actually just sent a note on kata-dev re: CI. | 19:40 |
| kata-dev-irc-bot | <eric.ernst> tl; dr - I want to know who the right person is for working with on adding CI or PPC64 | 19:41 |
| kata-dev-irc-bot | <eric.ernst> s/or/for | 19:41 |
| kata-dev-irc-bot | <niteshkonkar007> @eric.ernst: I shall be adding CI for ppc64le. | 19:45 |
| kata-dev-irc-bot | <niteshkonkar007> @eric.ernst: aah, not clear. You said "no" to moving to Jenkins ? | 19:49 |
| kata-dev-irc-bot | <niteshkonkar007> Oh I missed it. Where was it? | 19:51 |
| *** zerocoolback has quit IRC | 20:00 | |
| kata-dev-irc-bot | <eric.ernst> ok, great. | 20:04 |
| kata-dev-irc-bot | <niteshkonkar007> @eric.ernst: We use Jenkins or Travis today? Coz I see " continuous-integration/travis-ci/pr — The Travis CI build passed " when I submit a PR. | 20:14 |
| kata-dev-irc-bot | <eric.ernst> I had just sent it - kata-dev mailine list. | 20:15 |
| *** LinuxMe has quit IRC | 20:21 | |
| kata-dev-irc-bot | <eric.ernst> @jose.carlos.venegas.m @salvador.fuentes - I'm looking to formalize what I have at https://github.com/egernst/kata-deploy | 20:28 |
| kata-dev-irc-bot | <eric.ernst> I feel like packaging is the most-relevant location at this point, though I'm happy for input from folks. | 20:28 |
| *** LinuxMe has joined #kata-dev | 20:31 | |
| kata-dev-irc-bot | <eric.ernst> And, if it goes to packaging, perhaps some of the readme should be updated to describe all that is included in the repo (its getting bigger and has slightly different scope) | 20:31 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst yes packaging could be a good place but agree today packaging repository holds the release scripts , obs packaging and kernel (patches and config) bits. | 20:33 |
| kata-dev-irc-bot | <eric.ernst> I was debating there or runtime. | 20:34 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> We need some order in the repository | 20:34 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst in my mind packaging or its own repo sounds better | 20:36 |
| kata-dev-irc-bot | <eric.ernst> we have too many repos :slightly_smiling_face: | 20:36 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> hahaha | 20:37 |
| kata-dev-irc-bot | <eric.ernst> okay. Well, the next part is thinking about how we test today on K8S. | 20:37 |
| kata-dev-irc-bot | <eric.ernst> I think keeping a running cri-containerd and crio cluster makes sense. | 20:38 |
| kata-dev-irc-bot | <eric.ernst> and would save time. | 20:38 |
| kata-dev-irc-bot | <niteshkonkar007> @eric.ernst: ping (incase u missed the above comment) | 20:38 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst oh good you touch that topic - its begin a pain have just integration test for cri-containerd in the CI | 20:39 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> because of 2 thinks - the time I need to know if the CI pass or fails and the conflicts between configurations in CNI or crictl | 20:40 |
| kata-dev-irc-bot | <eric.ernst> Regardless, though, I want to maintain a released container image such as what is provided in kata-deploy repo today. | 20:40 |
| kata-dev-irc-bot | <eric.ernst> And then have that tested. | 20:40 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst we can publish it in the agent repo after each tag | 20:41 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> sorry - I thought the container.img and initrd | 20:42 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> you mean a dockerhub image? | 20:42 |
| kata-dev-irc-bot | <eric.ernst> I want to update the Dockerfile each time. | 20:47 |
| kata-dev-irc-bot | <eric.ernst> and make sure we pull in the right artifacts. | 20:47 |
| *** mcastelino has quit IRC | 20:49 | |
| kata-dev-irc-bot | <jose.carlos.venegas.m> ok ok | 20:49 |
| kata-dev-irc-bot | <eric.ernst> Both should be used. | 20:50 |
| kata-dev-irc-bot | <eric.ernst> @salvador.fuentes ^ | 20:50 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @salvador.fuentes I wonder if we cold use a jenkins job? | 20:50 |
| kata-dev-irc-bot | <eric.ernst> @jose.carlos.venegas.m the way I see it, we should just be updating the Dockerfile/triggering dockerhub @ same time as packaging is updated. | 20:51 |
| kata-dev-irc-bot | <eric.ernst> The main problem I foresee is that we have a different qemu being used. | 20:51 |
| kata-dev-irc-bot | <eric.ernst> I'm not sure having a 3rd built qemu in our repositories is something to strive for. | 20:52 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst yeah - add a 3rd build will increase our testing time more | 20:57 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> another option is add yet another job to test just the deploy with a static qemu - but is another set of CPUs we have to use | 20:59 |
| kata-dev-irc-bot | <sebastien.boeuf> @eric.ernst @jose.carlos.venegas.m @salvador.fuentes are you clear about what is done into our CI script ? And how we could simplify some things to speed it up ? | 20:59 |
| kata-dev-irc-bot | <sebastien.boeuf> if not, we should sync/talk about it | 21:00 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @sebastien.boeuf I dont' I can think that the kernel and qemu takes a long time to build | 21:00 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> there is an effort from @salvador.fuentes to install the packaged kernel if has not change | 21:01 |
| kata-dev-irc-bot | <sebastien.boeuf> @jose.carlos.venegas.m same for qemu ? | 21:01 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> not at the moment | 21:01 |
| kata-dev-irc-bot | <sebastien.boeuf> well those are the 2 pieces that we know are not changing a lot, and most of the time, we can use packages | 21:02 |
| kata-dev-irc-bot | <sebastien.boeuf> same for vm rootfs img | 21:02 |
| kata-dev-irc-bot | <sebastien.boeuf> is there a CRI-O pkg ? | 21:02 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> yes - actually I think crio takes some time to build as well | 21:03 |
| kata-dev-irc-bot | <sebastien.boeuf> yeah crio is a nightmare from a build time perspective | 21:03 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @sebastien.boeuf one problem with cri-o is that sometimes we need changes that are in master | 21:03 |
| kata-dev-irc-bot | <sebastien.boeuf> @jose.carlos.venegas.m which is fine if most of the time we can rely on pkgs | 21:04 |
| kata-dev-irc-bot | <salvador.fuentes> @sebastien.boeuf, @jose.carlos.venegas.m I was planning to work on qemu after the changes on the kernel script are merged | 21:04 |
| kata-dev-irc-bot | <sebastien.boeuf> @salvador.fuentes :slightly_smiling_face: | 21:04 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @sebastien.boeuf about the image I think does not take a long time but get some minutes from there too | 21:05 |
| kata-dev-irc-bot | <salvador.fuentes> @sebastien.boeuf also, the tests from CRI-O take around 5-7 min less minutes if they are run using a block device | 21:05 |
| kata-dev-irc-bot | <eric.ernst> @salvador.fuentes @jose.carlos.venegas.m - I'd like to treat the container iamge for kata-deploy as the default on how ppl use Kata in k8s | 21:05 |
| *** mcastelino has joined #kata-dev | 21:06 | |
| kata-dev-irc-bot | <eric.ernst> I'm happy for input here, but its how I see people using Kata in K8S easily, and I want to test it accordingly with this in mind (ie, our CI should make use of it) | 21:06 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst that sounds good to me | 21:06 |
| kata-dev-irc-bot | <eric.ernst> Then our CI as a first step for al lthese tests is "setup K8S + containerd" and "setup K8S + crio" | 21:06 |
| kata-dev-irc-bot | <eric.ernst> Or setup a cluster which does both. | 21:06 |
| kata-dev-irc-bot | <eric.ernst> Today we install and setup K8S each time we test, right? | 21:06 |
| kata-dev-irc-bot | <sebastien.boeuf> yes | 21:07 |
| kata-dev-irc-bot | <salvador.fuentes> yes | 21:07 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst I see have both in a same machine is problematic (cri-contaienrd and crio) | 21:07 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> we need to rework our CI to fix any conflict of having both | 21:07 |
| kata-dev-irc-bot | <eric.ernst> right, the cluster would need to be independent VMs | 21:07 |
| kata-dev-irc-bot | <eric.ernst> 1 for containerd, 1 for crio (doesn't make sense to mix) | 21:07 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst that simplify a lot the things | 21:07 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> and we can reuse the job for cri-containerd and cri-o use it as part of their testing | 21:08 |
| kata-dev-irc-bot | <salvador.fuentes> not sure that will work on Azure, as another level of virtualization would slow things a lot | 21:08 |
| kata-dev-irc-bot | <eric.ernst> no. | 21:09 |
| kata-dev-irc-bot | <eric.ernst> each one would be its own node. | 21:09 |
| kata-dev-irc-bot | <salvador.fuentes> ohh ok | 21:09 |
| kata-dev-irc-bot | <eric.ernst> Not 2 VMs on the node. | 21:09 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst @salvador.fuentes 2 jenkins jobs | 21:09 |
| kata-dev-irc-bot | <salvador.fuentes> ok, got it, yes, that should work | 21:09 |
| kata-dev-irc-bot | <eric.ernst> Or, 1 job that starts to VMs. | 21:09 |
| kata-dev-irc-bot | <eric.ernst> starts two VMs * | 21:09 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> horizontal scale the testing | 21:09 |
| kata-dev-irc-bot | <sebastien.boeuf> yes | 21:10 |
| kata-dev-irc-bot | <eric.ernst> ie, if we want to do more extensive testing, we should probably have a master/worker setup | 21:10 |
| *** LinuxMe has quit IRC | 21:10 | |
| kata-dev-irc-bot | <eric.ernst> Is there a 1:1 mapping of Jobs to azure VMs because of plugin today, or just how we are writing it? | 21:10 |
| kata-dev-irc-bot | <sebastien.boeuf> just to summarize, what are the big pieces that we currently build ? kernel, img, qemu, cri-o | 21:10 |
| kata-dev-irc-bot | <sebastien.boeuf> @salvador.fuentes what am I missing ? | 21:10 |
| kata-dev-irc-bot | <salvador.fuentes> runc, which I think we can get rid of | 21:11 |
| *** LinuxMe has joined #kata-dev | 21:12 | |
| kata-dev-irc-bot | <salvador.fuentes> also the cni binaries | 21:12 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> I think we install runc like 3 times | 21:12 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> one for docker | 21:12 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> one for cri-o | 21:12 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> and soon one for containerd | 21:12 |
| kata-dev-irc-bot | <salvador.fuentes> are all different versions? | 21:13 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> well, contaienrd install its own version (defined version in its setup ) | 21:13 |
| kata-dev-irc-bot | <salvador.fuentes> ohh, ok | 21:13 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> I think any version we use should be compatible with all our testing | 21:14 |
| kata-dev-irc-bot | <sebastien.boeuf> is that feasible to have packages for everything ? | 21:14 |
| kata-dev-irc-bot | <sebastien.boeuf> we know all the versions we want to use for those binaries, we can have them pre-buit | 21:14 |
| kata-dev-irc-bot | <sebastien.boeuf> built | 21:14 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @sebastien.boeuf for containerd no - but they provide static binaries | 21:15 |
| kata-dev-irc-bot | <sebastien.boeuf> oh nice | 21:15 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> that is just curl and tar -c / | 21:15 |
| kata-dev-irc-bot | <sebastien.boeuf> which is exactly what we need | 21:15 |
| *** mcastelino has quit IRC | 21:15 | |
| kata-dev-irc-bot | <sebastien.boeuf> the network from an Azure VM is super fast | 21:15 |
| *** mcastelino has joined #kata-dev | 21:15 | |
| kata-dev-irc-bot | <sebastien.boeuf> and uncompressing the tar is negligeable | 21:16 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @sebastien.boeuf oh yea - go is downloaded like in one second | 21:16 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> i love it | 21:16 |
| *** LinuxMe has quit IRC | 21:16 | |
| kata-dev-irc-bot | <eric.ernst> While I agree we should have all the prebuilts, etc, I think a lot of the problem in addition is that this is all serialized. | 21:17 |
| kata-dev-irc-bot | <sebastien.boeuf> but this whole thing would mean that for any binary not available online, we should generate them and make them available on a server somewhere | 21:17 |
| *** jdandrea has joined #kata-dev | 21:17 | |
| kata-dev-irc-bot | <jose.carlos.venegas.m> @eric.ernst @salvador.fuentes I can move this PR | 21:17 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> https://github.com/kata-containers/tests/pull/313/files | 21:17 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> to its own job | 21:17 |
| kata-dev-irc-bot | <jose.carlos.venegas.m> and then when we have kata deploy move that job to use it | 21:18 |
| kata-dev-irc-bot | <sebastien.boeuf> @eric.ernst definitely, but solving the overhead is a priority because if you run a 40min job with 30min setup, this is very annoying to think that 3/4 of the money/time spent is not even for the test | 21:18 |
| kata-dev-irc-bot | <sebastien.boeuf> @eric.ernst because in almost all cases, we need the same binaries, plus or minus some specific ones associated to the test you wanna run | 21:19 |
| kata-dev-irc-bot | <eric.ernst> (sorry spent the last 2 minutes looking for a "duh" emoticon) ;) | 21:19 |
| kata-dev-irc-bot | <sebastien.boeuf> :homer: | 21:19 |
| kata-dev-irc-bot | <sebastien.boeuf> lol | 21:19 |
| kata-dev-irc-bot | <sebastien.boeuf> @eric.ernst do you think this pool of binaries that we need could be stored somewhere ? Is that feasible ? | 21:22 |
| kata-dev-irc-bot | <eric.ernst> 'pool of binaries' | 21:22 |
| kata-dev-irc-bot | <sebastien.boeuf> they could be re-generated everytime we change a version | 21:23 |
| kata-dev-irc-bot | <eric.ernst> I think we should just manage which ones we already have | 21:23 |
| kata-dev-irc-bot | <eric.ernst> ie, we already have packages for qemu ,etc. | 21:23 |
| kata-dev-irc-bot | <eric.ernst> Use that. | 21:23 |
| kata-dev-irc-bot | <eric.ernst> no? | 21:23 |
| kata-dev-irc-bot | <eric.ernst> crio has a PPA, we just don't use / verify this. | 21:23 |
| kata-dev-irc-bot | <eric.ernst> And, tbh, I don't think CRIO takes that much time at this point anyway. We shouldn't do this on our side; we should make use of theirs. | 21:23 |
| kata-dev-irc-bot | <sebastien.boeuf> ok cool for crio | 21:23 |
| kata-dev-irc-bot | <sebastien.boeuf> ppa sounds okay to me | 21:24 |
| kata-dev-irc-bot | <eric.ernst> We have packages for all of our other components (rootfs, vmlinuz, qemu) | 21:24 |
| *** devimc has quit IRC | 21:25 | |
| kata-dev-irc-bot | <eric.ernst> for CRIO: https://launchpad.net/~projectatomic/+archive/ubuntu/ppa/+packages | 21:26 |
| kata-dev-irc-bot | <sebastien.boeuf> cool, so we just need to leverage them, right ? that's all that is missing ? | 21:27 |
| kata-dev-irc-bot | <eric.ernst> AFAIU. | 21:27 |
| kata-dev-irc-bot | <sebastien.boeuf> ok, now about a proper split for jobs, do you have something in mind ? | 21:28 |
| kata-dev-irc-bot | <sebastien.boeuf> better parallelization I mean | 21:28 |
| kata-dev-irc-bot | <eric.ernst> a few things come to mind: 1) we'd need more vCPUs 2) We'd need changes to jenkins to do this (which is what started updated discussions re: zuul) | 21:29 |
| kata-dev-irc-bot | <jonolson> @anne I think you started getting a Kata account set up on GCP — let’s see if we can finish that up and then I can ping our product folks (now that I’ve tracked some down) for a dose of credits | 21:31 |
| kata-dev-irc-bot | <eric.ernst> 3) I need to look @ what's in K8S today, and what is feasible. I"m sure plenty of time is spent installing and then bootstrapping k8s to only do a couple of superficial single-node tests. | 21:31 |
| kata-dev-irc-bot | <eric.ernst> which is maybe okay; but, this is where my focus is right now (specifically wanting to be able to test CRIO and Containerd using kata-deploy instead of (or in addition to) packages. | 21:32 |
| kata-dev-irc-bot | <eric.ernst> @jonolson - that'd be pretty interesting | 21:32 |
| kata-dev-irc-bot | <anne> @eric.ernst you _should_ have access to that account from when we set it up, but it’s been a while so I’m a little rusty on where we left it :) | 21:33 |
| kata-dev-irc-bot | <eric.ernst> I think we're both covered in rust and dust here. | 21:33 |
| kata-dev-irc-bot | <eric.ernst> I would need to first find where we talked about it, then find the archive for that medium of conversation :slightly_smiling_face: | 21:34 |
| kata-dev-irc-bot | <anne> Post release rust! I think actually it sent an email to your gmail when I added you as a user? | 21:38 |
| kata-dev-irc-bot | <anne> If that’s a dead end I can jump back in and resend in a bit here (or tomorrow) | 21:40 |
| clarkb | I've just responded to the zuul discussion thread fwiw. Currently we can't really do much until github is told to send us the web hook events (whcih I've asked for) | 21:41 |
| clarkb | but once that is in place we can start pushing on config updates and all that and show some concrete "this is what it looks like" | 21:42 |
| kata-dev-irc-bot | <sebastien.boeuf> thanks Clark ! | 21:43 |
| kata-dev-irc-bot | <eric.ernst> @sebastien.boeuf can you update github? I believe you have admin? | 21:49 |
| kata-dev-irc-bot | <sebastien.boeuf> @eric.ernst clarkb, I have just installed Zuul Github App to kata-containers/proxy | 21:51 |
| kata-dev-irc-bot | <sebastien.boeuf> anything else ? | 21:51 |
| clarkb | I think that gives us a place to start | 21:52 |
| *** corvus has joined #kata-dev | 21:52 | |
| clarkb | I'm still digging out of my summit hole so wouldn't expect the new job running this week. At least not if I am the one doing the config side. I've let the zuul folks know and maybe one of them will pick it up otherwise it will be high on my list of todos for next week | 21:53 |
| kata-dev-irc-bot | <sebastien.boeuf> sure, np | 21:53 |
| kata-dev-irc-bot | <sebastien.boeuf> but I didn't realize that's all you needed, I could have done that earlier | 21:53 |
| kata-dev-irc-bot | <sebastien.boeuf> anywa | 21:53 |
| corvus | i believe the app can be installed at the org level. we might want to go ahead and do that so we don't have to add it for each app? | 21:54 |
| kata-dev-irc-bot | <sebastien.boeuf> corvus, I have installed it only on the proxy for now, it's not a big deal to apply/install it for the whole org | 21:55 |
| kata-dev-irc-bot | <raravena80> imo, it would be a pain to build everything crio, containerd, qemu, etc for CI + test for many k8s versions. Maybe a simplified matrix would help. | 22:10 |
| kata-dev-irc-bot | <eric.ernst> File uploaded https://katacontainers.slack.com/files/U8A7TKV7G/FAYS5PZKM/screen_shot_2018-05-30_at_3.15.14_pm.png / https://slack-files.com/T86U7NQTT-FAYS5PZKM-2bab4bfc53 | 22:15 |
| kata-dev-irc-bot | <eric.ernst> ie: ^ | 22:15 |
| kata-dev-irc-bot | <eric.ernst> Depending on how often we need to really change the baseline, we could just say "setup a k8s cluster with 2 nodes, one using containerd, one using crio" | 22:16 |
| kata-dev-irc-bot | <eric.ernst> I guess problems areas wuld be if (1) we need to update the base versions often, (2) if using the booted cluster would be a bottleneck, (3) if it became expensive to leave this on. | 22:17 |
| kata-dev-irc-bot | <raravena80> would be nice to use 'spot' or 'prempted' instances. For updating the base images packer is a nice tool | 22:38 |
| kata-dev-irc-bot | <eric.ernst> was doing some reading, @salvador.fuentes @sebastien.boeuf - any particular rationale on the size of machine we use for kataCI? | 22:38 |
| kata-dev-irc-bot | <eric.ernst> This is what just runs jenkins, right? Seems we could do with a smaller machine if we wanted? | 22:38 |
| *** dlw has joined #kata-dev | 22:38 | |
| kata-dev-irc-bot | <salvador.fuentes> there is no particular reason, we decide to use it because we think it was enough, and its price is not high | 22:39 |
| kata-dev-irc-bot | <salvador.fuentes> but if we need more power, we can change it | 22:39 |
| kata-dev-irc-bot | <sebastien.boeuf> @eric.ernst the size of the VM for Jenkins is what was suggested by Axure | 22:39 |
| kata-dev-irc-bot | <sebastien.boeuf> Azure | 22:39 |
| kata-dev-irc-bot | <eric.ernst> I looked at last month of usage; one time we hit 40% cpu util :slightly_smiling_face: | 22:40 |
| kata-dev-irc-bot | <eric.ernst> ok. | 22:40 |
| kata-dev-irc-bot | <eric.ernst> If I was selling you a service, I'd suggest a bigger machine ;) | 22:40 |
| kata-dev-irc-bot | <sebastien.boeuf> since we started from the Jenkins image proposed by Azure market place | 22:40 |
| kata-dev-irc-bot | <eric.ernst> ok | 22:40 |
| kata-dev-irc-bot | <eric.ernst> makes sense | 22:40 |
| kata-dev-irc-bot | <sebastien.boeuf> I agree, but I think it is very inexpensive, right ? | 22:40 |
| kata-dev-irc-bot | <sebastien.boeuf> somthing like 70 per month | 22:40 |
| kata-dev-irc-bot | <eric.ernst> sure. you could spend 8$ a month too. I was just reading into it trying to see where to manage vCPUs | 22:41 |
| kata-dev-irc-bot | <raravena80> spot is hella cheap on AWS, like 1/4th of the price. But, they don't have nested virt support yet, afaik. | 22:43 |
| kata-dev-irc-bot | <sebastien.boeuf> there are some 8$ per month machine ? | 22:43 |
| kata-dev-irc-bot | <sebastien.boeuf> @raravena80, we don't need the virt support for Jenkins VM | 22:43 |
| kata-dev-irc-bot | <eric.ernst> yes. | 22:43 |
| kata-dev-irc-bot | <sebastien.boeuf> this is simply the master node | 22:43 |
| kata-dev-irc-bot | <raravena80> we probably don't want to put the master in a spot instance | 22:43 |
| kata-dev-irc-bot | <raravena80> more for the slaves | 22:44 |
| kata-dev-irc-bot | <sebastien.boeuf> @eric.ernst don't worry about it, soon we won't need Jenkins :slightly_smiling_face: | 22:44 |
| kata-dev-irc-bot | <sebastien.boeuf> Zuul will be hosted somewhere for us | 22:44 |
| kata-dev-irc-bot | <sebastien.boeuf> ;) | 22:44 |
| kata-dev-irc-bot | <eric.ernst> :slightly_smiling_face: | 22:49 |
| kata-dev-irc-bot | <eric.ernst> I was going to ask for more vCPUs -- I see we have a lot on the east coast already though. | 22:49 |
| kata-dev-irc-bot | <eric.ernst> (210). | 22:49 |
| kata-dev-irc-bot | <eric.ernst> And more on the west coast as well. | 22:49 |
| kata-dev-irc-bot | <eric.ernst> @salvador.fuentes @sebastien.boeuf - how are we deciding where to send jobs? | 22:50 |
| kata-dev-irc-bot | <eric.ernst> on a per job basis, or? | 22:50 |
| kata-dev-irc-bot | <salvador.fuentes> ohh, no, right now we are sending all of them to east | 22:50 |
| kata-dev-irc-bot | <salvador.fuentes> because the jenkins plugin does not allow us to have 2 different pools of resources | 22:51 |
| kata-dev-irc-bot | <eric.ernst> Because that's where 80% of the vCPUs are available? | 22:51 |
| kata-dev-irc-bot | <eric.ernst> ok. | 22:51 |
| kata-dev-irc-bot | <salvador.fuentes> or at least I couldnt find a way to have them on both zones | 22:51 |
| kata-dev-irc-bot | <eric.ernst> Did you ask for a bigger quota before for vCPUs? | 22:51 |
| kata-dev-irc-bot | <salvador.fuentes> yes | 22:51 |
| kata-dev-irc-bot | <eric.ernst> 210 sounded like a funny number | 22:51 |
| kata-dev-irc-bot | <salvador.fuentes> ohh, it was because we have 10 by default and asked for 200 more | 22:52 |
| kata-dev-irc-bot | <eric.ernst> 10! | 22:52 |
| kata-dev-irc-bot | <eric.ernst> that must've went pretty quick. | 22:52 |
| kata-dev-irc-bot | <salvador.fuentes> and I originally asked for more on west, but they said that they didnt have enough capacity, so thats why I asked for more on the east | 22:53 |
| kata-dev-irc-bot | <salvador.fuentes> yeap, we had 10 at the very beginning, but you could only launch like 2 vms in parallel | 22:53 |
| kata-dev-irc-bot | <eric.ernst> @jose.carlos.venegas.m - can you help me understand why your CI PR has containerd built from source? | 23:00 |
| *** sameo has quit IRC | 23:06 | |
| *** LinuxMe has joined #kata-dev | 23:07 | |
| *** LinuxMe has quit IRC | 23:11 | |
| *** mrcastelnix has quit IRC | 23:41 | |
| *** zerocoolback has joined #kata-dev | 23:43 | |
| *** dlw has quit IRC | 23:53 | |
| *** zerocoolback has quit IRC | 23:57 | |
| *** zerocoolback has joined #kata-dev | 23:59 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!
