| *** os-chef-bot has quit IRC | 03:12 | |
| *** os-chef-bot has joined #openstack-chef | 03:12 | |
| *** nkrinner_afk is now known as nkrinner | 07:12 | |
| *** jojden has joined #openstack-chef | 07:29 | |
| *** furlongm has quit IRC | 08:30 | |
| *** furlongm has joined #openstack-chef | 08:31 | |
| *** Damjanek has quit IRC | 09:26 | |
| *** Damjanek has joined #openstack-chef | 09:26 | |
| *** eroux has joined #openstack-chef | 12:30 | |
| *** eroux has quit IRC | 12:37 | |
| *** eroux has joined #openstack-chef | 12:38 | |
| *** mutin-sa has joined #openstack-chef | 13:24 | |
| *** mutin-s has quit IRC | 13:27 | |
| *** laurent\_ has joined #openstack-chef | 13:30 | |
| *** laurent\ has quit IRC | 13:32 | |
| *** jperry has joined #openstack-chef | 14:14 | |
| sc` | i have what i think is a good start to two-node gates, but without pushing it to gerrit, i have no way of knowing if what i write is even going to work | 14:17 |
|---|---|---|
| sc` | frickler: would you be in a position to review such things? | 14:23 |
| *** eroux has quit IRC | 14:43 | |
| *** eroux has joined #openstack-chef | 14:46 | |
| frickler | sc`: maybe, but it will mostly still be try-and-error. I've discussed this with jklare earlier today, and we are not sure whether it will be worth the effort. we are assuming that going multi-node will open up a new can of worms for sure, mainly networking related which we can completely ignore for aio. doing that in parallel to dealing with OOM issues seems bad | 14:47 |
| frickler | sc`: the alternative would be that maybe you could consider setting up a 3rd party CI system like we did here | 14:48 |
| sc` | i have no such resources at my disposal :/ | 14:48 |
| sc` | that's partly why things have dragged on for so long | 14:48 |
| sc` | it wasn't until talking with people and reading code did it become apparent that multinode was a thing. repo has the memory limits for the controller at 4G, so i thought that piggy could run on two nodes | 14:50 |
| sc` | it's only allinone that is our problem | 14:50 |
| sc` | just don't advocate for allinone | 14:51 |
| sc` | if that doesn't make sense, or sounds like a Bad Idea, do let me know | 14:52 |
| frickler | sc`: I don' think it makes sense to setup multi-node on infra without having good support from an infra-root, allowing for properly debugging the networking issues that will come up | 14:53 |
| frickler | sc`: maybe we could first give it a try in our local CI env, but you'd have to convince jklare that that would be a good idea, first | 14:54 |
| frickler | sc`: other option would be trying to get a different sponsor for a second 3rd party CI. we wouldn't need more than a small controller and a quota of a couple of 12G instances. we could even do with a quota of 1, if needed | 14:57 |
| sc` | of course. i'm doing this on my own time. i wouldn't expect anyone to assist unless it was interesting and worth the time | 14:57 |
| frickler | sc`: so workday isn't interested in this anymore? | 14:58 |
| sc` | they're still interested in their own way | 14:58 |
| sc` | particularly, once mitaka is eol, i'd expect someone to poke their head in | 14:58 |
| frickler | sc`: another idea that just comes to my mind, maybe we could try to do a some-in-one, like leave out ceilometer and heat for a start, and see how far that will work? | 15:04 |
| sc` | possible. i did see that puppet disables some services depending on what they're building | 15:04 |
| sc` | that, and tripleo, have been my inspiration for this multinode madness | 15:05 |
| sc` | apparently this is a problem all over | 15:05 |
| sc` | solving it for chef would help the remaining five openstack-chef users ;) | 15:06 |
| frickler | sc`: yes, and it seems to be even more severe with ocata, once we do the switch | 15:07 |
| sc` | one solution on the infra side would be doubling the size of single use instances, that was mentioned when i was chatting with them, but that comes with a huge cost | 15:08 |
| frickler | sc`: yes, I've seen that. trouble is that they only have one pool of instances. not sure whether zuulv3 would be more flexible in the future | 15:10 |
| sc` | yup. that's more the nuclear option. the options for us would be to help put openstack itself on a diet, or ci work | 15:11 |
| sc` | we could split resources across clouds like tripleo does, but that would be getting buy-in from them | 15:12 |
| sc` | i think tripleo is using rhcloud, so there'a completely different relationship there | 15:13 |
| frickler | the more hooks we need in order to setup testing in a special way, the less useful will it be, because it will be different from what people deploy. we are working locally on a heat stack that will mimic our real world deployment, with multiple interfaces per node, that would IMHO be a good solution for a proper CI setup | 15:17 |
| sc` | of course. i didn't intend on reinventing the whole of tripleo ci for chef, but leveraging the useful parts like what nodepool exposes | 15:21 |
| -openstackstatus- NOTICE: We are currently investigating an issue with our AFS mirrors which is causing some projects jobs to fail. We are working to correct the issue. | 15:48 | |
| *** eroux has quit IRC | 16:41 | |
| *** jojden has quit IRC | 16:50 | |
| *** nkrinner is now known as nkrinner_afk | 17:18 | |
| -openstackstatus- NOTICE: AFS replication issue has been addressed. Mirrors are currently re-syncing and coming back online. | 17:18 | |
| *** jperry has quit IRC | 17:23 | |
| *** jperry has joined #openstack-chef | 17:24 | |
| *** Edi_ has joined #openstack-chef | 17:41 | |
| Edi_ | Guide Deploying Openstack with Chef? | 17:43 |
| Edi_ | Hi Guys | 17:43 |
| Edi_ | Guide install Openstack with Chef | 17:43 |
| Edi_ | where to find? | 17:43 |
| *** Edi_ has quit IRC | 18:27 | |
| *** eroux has joined #openstack-chef | 18:43 | |
| *** emagana has joined #openstack-chef | 18:54 | |
| sc` | ^ that is a problem :/ | 19:14 |
| sc` | could topic link to jklare's post from mitaka. it's still relevant | 19:15 |
| *** jperry has quit IRC | 22:55 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!