| *** sdake has quit IRC | 00:03 | |
| cloudnull | There were hash problems earlier today but I'm not sure at this point. I can help dig into it in a little while. | 00:11 |
|---|---|---|
| openstackgerrit | Merged stackforge/os-ansible-deployment: Add newline to /etc/ssh/sshd_config if necessary https://review.openstack.org/176758 | 00:13 |
| *** sdake has joined #openstack-ansible | 01:22 | |
| *** sdake_ has quit IRC | 01:26 | |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Allow users to configure the Glance flavor in use https://review.openstack.org/176464 | 01:42 |
| *** sdake_ has joined #openstack-ansible | 01:50 | |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Updates for tag 10.1.4 https://review.openstack.org/177050 | 01:51 |
| *** sdake has quit IRC | 01:53 | |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Update for tag 9.0.8 https://review.openstack.org/177051 | 01:55 |
| *** JRobinson__ has quit IRC | 02:07 | |
| *** devlop has joined #openstack-ansible | 02:39 | |
| *** galstrom_zzz is now known as galstrom | 02:47 | |
| *** devlop has quit IRC | 03:00 | |
| *** devlop has joined #openstack-ansible | 03:00 | |
| *** sdake has joined #openstack-ansible | 03:00 | |
| *** galstrom is now known as galstrom_zzz | 03:02 | |
| *** javeriak has quit IRC | 03:03 | |
| *** sdake_ has quit IRC | 03:04 | |
| *** sdake_ has joined #openstack-ansible | 03:04 | |
| *** bilal_ has quit IRC | 03:06 | |
| *** sdake has quit IRC | 03:08 | |
| *** galstrom_zzz is now known as galstrom | 03:48 | |
| *** devlop has quit IRC | 04:00 | |
| *** galstrom is now known as galstrom_zzz | 04:10 | |
| *** subscope_ has joined #openstack-ansible | 04:30 | |
| *** JRobinson__ has joined #openstack-ansible | 04:45 | |
| *** markvoelker_ has quit IRC | 04:51 | |
| *** subscope_ has quit IRC | 06:11 | |
| *** stevemar has quit IRC | 06:32 | |
| *** JRobinson__ has quit IRC | 07:05 | |
| *** sdake_ has quit IRC | 07:20 | |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add release pinning to galera roles https://review.openstack.org/176720 | 08:03 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Make script executable and BSD sed compatible https://review.openstack.org/177115 | 08:07 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Make script executable and BSD sed compatible https://review.openstack.org/177115 | 08:19 |
| *** javeriak has joined #openstack-ansible | 08:32 | |
| *** javeriak has quit IRC | 09:12 | |
| svg | can anyone give me some hints on the disk space sizing for metal hosts? I assume logging hosts (elasticsearch) and especially infra hosts (database) are specific here | 09:13 |
| svg | seems I can't find any hints on how big mysql could grow | 09:13 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Set external roles to import into ansible namespace https://review.openstack.org/177131 | 09:14 |
| odyssey4me | svg it all depends on how big your openstack environment will grow, but any sizing recommendations that you'll find for openstack in general would apply | 09:15 |
| svg | well, thats the thing, I cant find anythong that gives even the wildest guess/idea. | 09:16 |
| mattt | odyssey4me: do deleted records now actually removed from the dbs? | 09:16 |
| odyssey4me | you've identified the two main ones, the only other one is where the glance container lives due to the fact that by default we set glance to cache images - you can override it though and you can also set your glance container sizing appropriately | 09:16 |
| svg | hm, given we use ceph, I think I disable glance caching | 09:17 |
| svg | or at least it should be | 09:17 |
| odyssey4me | mattt honestly I have no idea - I know that since Havana (or was it Icehouse) there is a set of shadow tables which acts as some sort of revert archive for the projects | 09:17 |
| odyssey4me | svg ah yes, I think I saw you doing that in your code :) | 09:17 |
| svg | deploying an image to a vulume is a pure ceph thing | 09:17 |
| svg | basically, I want to know how many disk space I should provision on the infrahosts, at least enough to get us started for a while | 09:18 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Set external roles to import into ansible namespace https://review.openstack.org/177131 | 09:18 |
| svg | upstream openstack mentioned avuelt '100gb' for a controller node, but that's the best I could find | 09:18 |
| svg | .. mentioned a vague.. | 09:19 |
| odyssey4me | svg I'm not sure what defaults we use at RAX, but in most cases I expect that the typically lowest size SAS drive of 160GB will do as a starting point? | 09:19 |
| odyssey4me | I have no empirical data to work from though. | 09:19 |
| odyssey4me | You may wish to have something like 60GB for your OS and 100GB for your lxc VG. Test that and see how it goes? | 09:20 |
| svg | odyssey4me: we have blades with one ssd of 75GB, so we plan to mount a ceph block device on /var/lib/lxc/ | 09:21 |
| svg | so yes, at least 100gb was the idea for now | 09:22 |
| odyssey4me | svg in my own experience I found that using a ceph block device in a block like that wasn't very reliable... but that experience is around a year old | 09:22 |
| odyssey4me | issues we found were: | 09:23 |
| svg | I have a metal host with around 22 containers taking at fresh deploy around 30gb | 09:23 |
| svg | (infra+storage host combined) | 09:23 |
| odyssey4me | 1) resizing that block device was a bit of a pain, it is definitely better to have a block device per instance to simplify resizing when need be so that your down-time isn't for the whole controller but instead just for the instance that needs a resize | 09:24 |
| svg | hm, goven the random container names, that moight be a pain to provision | 09:24 |
| svg | but good point | 09:24 |
| odyssey4me | 2) when you reboot the host with a block device mount, sometimes it causes a kernel panic and other times it causes an unclean shutdown... you have to very manually shut down each container before manually dismounting the device (which is when the kernel panic happens), then reboot the host | 09:25 |
| odyssey4me | 3) on boot up the mounting of the block device doesn't happen properly because it's not a normal block device... you may need to set the containers not to start automatically on boot and you may also have to remount the block device manually | 09:26 |
| odyssey4me | to be fair though, this is the experience based on Ceph Calamari, I think the code-name was | 09:26 |
| odyssey4me | YMMV based on newer Ceph versions | 09:27 |
| svg | 2 and 3 sounds very to the point | 09:27 |
| svg | not sure why resizing was a pain though | 09:27 |
| odyssey4me | a pain for a few reasons: | 09:27 |
| odyssey4me | 1) the resize operation when I was doing it was not an online thing - you had to shut the containers down, take the pool off-line, copy the pool to a new one that's larger, remove the old pool, set the new pool to be mounted in place of the old one, then mount it and bring the containers online again | 09:29 |
| odyssey4me | oh, and of course you had to extend the partitions inside the containers, or something like that | 09:30 |
| odyssey4me | not sure if that's still the process - check that out as it's an important operational process to be aware of when using ceph for your controllers | 09:30 |
| svg | afaik, resizing an rbd is just one command now | 09:31 |
| odyssey4me | ok, is it online | 09:32 |
| odyssey4me | bear in mind that all it'll do is increase the available disk - you still then need to do partition and file system expansions | 09:32 |
| svg | I'm more worries about your 2/ and 3/ though | 09:33 |
| odyssey4me | lvm should make that easier | 09:33 |
| odyssey4me | svg yeah, those were a pain - generally in that environment we resolved to rather use local disks for controllers to maintain reliability and seperation of concerns | 09:34 |
| svg | sure, but that's pretty straightfoward | 09:34 |
| svg | btw, kilo release is imminent right? does osad follow that quickly? | 09:34 |
| odyssey4me | within openstack there weren't issues as it was one rbd image per volume and the attach/detach/reattach was reliable for the most part | 09:34 |
| odyssey4me | svg yeah, kilo is in RC2 I think and we've already got QE teams testing based on the latest kilo patch sets... so the osad kilo release will likely follow very quickly after upstream kilo releases | 09:35 |
| odyssey4me | we're just busy fighting some changes to Horizon dependencies upstream which appear to have broken things... busy tracking that down and figuring out how we can get past it | 09:36 |
| *** sdake has joined #openstack-ansible | 09:37 | |
| odyssey4me | mattt any luck tracking down a fix for the Horizon issue - or are you chasing the master/heat squirrel? | 09:37 |
| svg | ok, so now I'm still working on a juno setup, with my ceph patch, but given we hope do to a production setup by the end of May, that probably will be based on kilo | 09:38 |
| svg | which also means, I'll need to reimplement the ceph patch on kilo, and I guess that will be closer to upstream it first on master, and backport it in kilo | 09:39 |
| *** sdake_ has joined #openstack-ansible | 09:39 | |
| odyssey4me | svg yeah, that's the usual workflow | 09:39 |
| svg | which means I can work on it more seriously in the coming weeks | 09:40 |
| odyssey4me | once we release kilo we'll also start working on implementing Ceph in master - we're targeting that for a feature release so we'll likely be collaborating on doing that at the same time | 09:40 |
| odyssey4me | I'm not sure of the exact timelines, so you may end up doing the first draft before we dig into it - but your work so far has been a really good start! | 09:41 |
| svg | I didn't get much feedback on the ceph WIP patch - my primary concern was to know if the style of roles/playbooks/vars is in line with what is expected | 09:42 |
| *** sdake has quit IRC | 09:43 | |
| odyssey4me | svg yeah, we've been very busy focusing on kilo so we haven't had much time to go through it properly | 09:43 |
| odyssey4me | from what I've seen and some off-the-cuff discussions we had the style is good in general, but yeah we haven't had much time to dig into it | 09:44 |
| svg | I probably will on an updated kilo patch, as one of the things we want to test in coming weeks is upgrading the stack, so with juno>kilo now is a good time | 09:44 |
| odyssey4me | sorry about that, but you know how resourcing priorities work - we all have deadlines and yours and ours aren't quite matching just yet :/ | 09:45 |
| svg | no need to apologise, just wanted to know how things get going, and how I can optimize my private work to show it upstream | 09:45 |
| odyssey4me | svg we have this to help with upgrading: https://github.com/stackforge/os-ansible-deployment/blob/kilo/scripts/run-upgrade.sh | 09:46 |
| odyssey4me | that'll be something to consider as part of your upgrade testing and kilo patch | 09:46 |
| mattt | odyssey4me: i don't think we need scss | 09:47 |
| mattt | odyssey4me: looks like it should be using pyScss instead | 09:47 |
| odyssey4me | mattt oh? this is the error I'm getting: django_pyscss.compressor.DjangoScssFilter: not found | 09:47 |
| mattt | odyssey4me: which version of django-scss are you using? | 09:48 |
| odyssey4me | aha, I see it - let me try that out | 09:48 |
| mattt | i've not gotten that far, but i just noticed that scss isn't even in global-requirements.txt | 09:48 |
| odyssey4me | yup, I think you may be right | 09:48 |
| svg | that script looks very throrough. the process still sounds scary though :) | 09:49 |
| odyssey4me | svg it's not that bad - openstack upgrades are pretty good these days | 09:51 |
| svg | openstack in general or thanks to osad? | 09:51 |
| odyssey4me | in the old days with essex and before it was nasty, but the upgrade process got to be a lot more reliable from around Havana I think it was | 09:51 |
| odyssey4me | openstack in general | 09:52 |
| svg | ok, so most people I hear talking ugly stories about this, are talking aout older experiences I guess | 09:52 |
| odyssey4me | osad, well, we've done the script because the v10/juno to the v11/kilo upgrade involves a few architecture changes and the script takes care of doing that all for you | 09:52 |
| odyssey4me | svg yeah, most of the time the upgrades are painful due to fiddles people have done to fix bugs they found which weren't backported to whatever release they were using... so you have to review code, etc | 09:53 |
| odyssey4me | with osad we're deploying from git sources so you could, in theory, have your own git repo with openstack source and deploy from there | 09:54 |
| svg | uhuh | 09:54 |
| odyssey4me | so your bugfixes can go through a normal development process and you can throw developers against it instead of throwing operations guys at it | 09:54 |
| odyssey4me | alternatively you could just track the upstream repo at whatever level you want - whether it be a tagged version, the 'stable/blah' branch, or a sha which has the fix you need | 09:55 |
| odyssey4me | osad makes it flexible to do that | 09:55 |
| svg | for now, given our limited experience, we'll stick to the well tested upstream version :) | 09:56 |
| odyssey4me | yup, with great power comes great responsibility :) | 09:58 |
| svg | ow, I see e.g. the mysql db folder in the container is mounted separately from /opemstack/... | 10:00 |
| odyssey4me | mattt I removed scss and ensure that pyScss was installed, but no go | 10:00 |
| svg | so extra disk space is more needed on that mount point | 10:00 |
| odyssey4me | svg yeah the mysqldb folder it a bind mount to the host - different to most of the others | 10:01 |
| odyssey4me | the idea there is that the container can be replaced, but the data folder will stay behind... also you can backup the data using a backup agent on the host | 10:01 |
| svg | oh, and there are even daily backups | 10:03 |
| mattt | odyssey4me: what version of django-pyscss ? | 10:06 |
| mattt | needs to be <2.0.0 | 10:06 |
| *** markvoelker has joined #openstack-ansible | 10:06 | |
| odyssey4me | django-compressor==1.5 | 10:06 |
| odyssey4me | django-pyscss==1.0.6 | 10:06 |
| odyssey4me | pyScss==1.2.1 | 10:06 |
| odyssey4me | XStatic-Bootstrap-SCSS==3.2.0.0 | 10:06 |
| mattt | that all looks ok, hmm | 10:06 |
| mattt | but no django-pyscss | 10:07 |
| mattt | odyssey4me: https://github.com/openstack/horizon/blob/ea92edacbb3daf77326549f4b6b61e210068ebba/requirements.txt#L17 | 10:07 |
| odyssey4me | there is - second line I posted | 10:07 |
| mattt | derp | 10:07 |
| * mattt drinks more coffee | 10:09 | |
| *** markvoelker has quit IRC | 10:11 | |
| odyssey4me | the class it's looking for definitely exists... grrrr | 10:12 |
| mattt | odyssey4me: yeah seeing that now | 10:15 |
| odyssey4me | stevelle stable/kilo's horizon does not complete our play due to the execution of 'horizon-manage.py compress --force' failing | 10:21 |
| odyssey4me | the error is CommandError: An error occured during rendering /usr/local/lib/python2.7/dist-packages/openstack_dashboard/templates/_stylesheets.html: /bin/sh: 1: django_pyscss.compressor.DjangoScssFilter: not found | 10:22 |
| mattt | Compressing... CommandError: An error occured during rendering /usr/local/lib/python2.7/dist-packages/openstack_dashboard/templates/_stylesheets.html: /bin/sh: 1: django_pyscss.compressor.DjangoScssFilter: not found | 10:29 |
| mattt | /bin/sh ?? | 10:29 |
| odyssey4me | mattt interesting: sh -> dash | 10:31 |
| openstackgerrit | Andy McCrae proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177160 | 10:35 |
| stevelle | django-pyscss released a 2.0.1 since yesterday | 10:38 |
| stevelle | interesting | 10:38 |
| openstackgerrit | Andy McCrae proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177160 | 10:42 |
| openstackgerrit | Andy McCrae proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177163 | 10:45 |
| odyssey4me | stevelle mattt busy in #openstack-horizon trying to track the issue down | 10:47 |
| mattt | odyssey4me: following along :) | 10:49 |
| *** sdake_ has quit IRC | 10:53 | |
| openstackgerrit | Andy McCrae proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177172 | 10:58 |
| odyssey4me | mattt stevelle how do we generate a fresh example local_settings.py file from source? | 11:00 |
| mattt | odyssey4me: ansible creates taht | 11:01 |
| mattt | there is a .example in /usr/local/lib/python2.7/dist-packages/openstack_dashboard/local | 11:02 |
| *** markvoelker has joined #openstack-ansible | 11:07 | |
| mattt | looks like we're still configuring horizon w/ a db even tho the db is no longer getting created | 11:09 |
| mattt | at least, i'm assumign that's why it's failing to connect to it :) | 11:10 |
| stevelle | I'm still kinda stumped | 11:11 |
| *** markvoelker has quit IRC | 11:12 | |
| odyssey4me | mattt oh? that and https://github.com/openstack/horizon/blob/stable/kilo/openstack_dashboard/settings.py#L27 appears to be missing from our local_settings.py template | 11:12 |
| stevelle | https://review.openstack.org/#/c/170988/ | 11:16 |
| stevelle | we need to get that change in I think | 11:17 |
| stevelle | for other reasons | 11:17 |
| odyssey4me | stevelle in the testing I'm doing we're using sha's from this morning | 11:18 |
| odyssey4me | but, as you've mentioned - since we prepared our templates there have been updates and we need to review those | 11:18 |
| mattt | odyssey4me: https://gist.github.com/mattt416/58fd9314f947d4ee0d4d | 11:18 |
| openstackgerrit | Andy McCrae proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177172 | 11:18 |
| openstackgerrit | Andy McCrae proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177160 | 11:20 |
| stevelle | odyssey4me: the exceptions on lines 77-79 of local_settings.py.example from that patch | 11:20 |
| stevelle | guessing that would be unrelated though | 11:20 |
| odyssey4me | stevelle prepa a patch for master while you're there :) | 11:21 |
| openstackgerrit | Andy McCrae proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177163 | 11:22 |
| odyssey4me | cloudnull ping when you're online, we need to track down which package is pulling in scss as a dependency as we're not explicitly pulling it in | 11:45 |
| Sam-I-Am | stevelle: isnt it kind of early there? | 11:49 |
| stevelle | um, yup | 11:49 |
| Sam-I-Am | less work, moar zzz? | 11:49 |
| stevelle | I might just start happy hour early | 11:50 |
| Sam-I-Am | i'm planning that for 9 am | 11:50 |
| mattt | stevelle: i want to visit portland, maybe i can go to the next pycon | 11:50 |
| mattt | even tho i don't py | 11:50 |
| stevelle | I look forward to it mattt. | 11:50 |
| Sam-I-Am | mattt: but pie is so tasty | 11:51 |
| Sam-I-Am | odyssey4me: whats this issue? | 11:51 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add release pinning to galera roles https://review.openstack.org/176720 | 11:52 |
| odyssey4me | Sam-I-Am so regardless of the fact that we don't ask for the python module scss to be around, it ends up on the horizon containers - and it and pyScss conflict in name spaces | 11:53 |
| odyssey4me | so we get a failure when trying to do the css compression | 11:53 |
| mattt | odyssey4me: nuked horizon containers, rebuilt, ran os-horizon.yml, and everything installed fine | 11:53 |
| odyssey4me | mattt I've revised the patch and updated the sha's as well for good measure | 11:53 |
| odyssey4me | let's see how it looks | 11:54 |
| mattt | sweet | 11:54 |
| Sam-I-Am | interesting | 11:54 |
| odyssey4me | I'm going to rebuild my containers as well | 11:54 |
| odyssey4me | actually, I'll rebuild the whole AIO | 11:54 |
| *** markvoelker has joined #openstack-ansible | 11:54 | |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Make script executable and BSD sed compatible https://review.openstack.org/177115 | 11:55 |
| *** mnestheu1 has joined #openstack-ansible | 12:02 | |
| *** galstrom_zzz is now known as galstrom | 12:06 | |
| cloudnull | mornings | 12:06 |
| cloudnull | odyssey4me: ping | 12:07 |
| cloudnull | just read scroll back. whats the word on scss and horizon broken for all the things ? | 12:07 |
| *** subscope has quit IRC | 12:11 | |
| odyssey4me | cloudnull we're testing now, we've definitely found that scss must not be on the horizon containers - the right python package is pyScss | 12:12 |
| cloudnull | ah thats exciting | 12:13 |
| odyssey4me | is there a way to determine on the repo container why a package was included? | 12:13 |
| odyssey4me | unfortunately both packages use the same namespace, so they conflict if they're both there | 12:13 |
| odyssey4me | you have to literally remove both, then add pyScss back | 12:13 |
| odyssey4me | haha: from https://pypi.python.org/pypi/scss/ | 12:14 |
| odyssey4me | Warning: Im not support python-scss anymore. Use pyscss by Kronus. | 12:14 |
| cloudnull | so this makes a lot of sense, i was looking back over logs yesterday and I had http://cdn.pasteraw.com/3jmmaktxjg2zmqn2atvxep2apr01xv8 as a diff for a broken kilo and a working master system. | 12:17 |
| cloudnull | odyssey4me yes, you can look at the report | 12:17 |
| cloudnull | which has all the packages and what was listing them as a dep. | 12:17 |
| cloudnull | IE something like this http://104.130.161.213:8181/reports/release-kilo-report.json | 12:18 |
| odyssey4me | ok, good news - with pyScss now in our requirements the older defunct scss is no longer being pulled in | 12:18 |
| odyssey4me | still verifying whether we get a successful build completion | 12:18 |
| cloudnull | in reading the report anything thats under "_requirements_" is user defined | 12:19 |
| cloudnull | which i have "scss" as being user defined (because we were adding it to the PR for kilo) and "pyScss" as a dep for horizon. | 12:19 |
| odyssey4me | cloudnull yeah, I've updated the review to switch to the right package | 12:20 |
| cloudnull | nice | 12:20 |
| * cloudnull fingers crossed | 12:20 | |
| odyssey4me | see https://review.openstack.org/176720 which is around half-way through the gate check | 12:20 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Add basic orchestration tests to the gate check https://review.openstack.org/176166 | 12:22 |
| *** andymccr has quit IRC | 12:24 | |
| cloudnull | well thats exciting http://logs.openstack.org/60/177160/3/check/os-ansible-deployment-dsvm-check-commit/e6ea45c/console.html | 12:25 |
| *** subscope has joined #openstack-ansible | 12:25 | |
| mattt | cloudnull: oh boy | 12:27 |
| cloudnull | all console logs are throwing 5xx | 12:27 |
| Sam-I-Am | you didnt need them anyway | 12:27 |
| mattt | openstackgerrit: you shouldn't have to specify pyScss | 12:27 |
| mattt | ^^^ odyssey4me | 12:28 |
| mattt | it'll get pulled in via requirements | 12:28 |
| *** galstrom is now known as galstrom_zzz | 12:28 | |
| mattt | horizon has a proper dependency on it | 12:28 |
| odyssey4me | mattt alright, I can set that - has pathlib also been set properly? | 12:28 |
| odyssey4me | it doesn't appear so | 12:29 |
| mattt | odyssey4me: i wanted to test if we needed that explicitly, not had a chance yet :( | 12:29 |
| cloudnull | asked in infra | 12:30 |
| mattt | odyssey4me: let's see if this guy builds correctly before we restart the build | 12:30 |
| mattt | currently doing neutron | 12:30 |
| odyssey4me | mattt fair enough :) | 12:30 |
| odyssey4me | pathlib doesn't appear to be anywhere | 12:30 |
| odyssey4me | should that be a horizon bug perhaps? | 12:31 |
| mattt | i think it was, sec | 12:31 |
| mattt | finding it | 12:31 |
| odyssey4me | https://bugs.launchpad.net/devstack/+bug/1447476 | 12:31 |
| openstack | Launchpad bug 1447476 in OpenStack Dashboard (Horizon) "No module named pathlib" [Critical,Fix released] | 12:31 |
| mattt | that's it! | 12:32 |
| mattt | so caused by django-pyscss 2.0.1 | 12:32 |
| mattt | now we're back down to < 2.0.0 it shouldn't be needed | 12:32 |
| odyssey4me | yeah, it appears that it should be pulled in from django-pyscss | 12:33 |
| odyssey4me | so the updated sha's should take care of it - but let's wait and see the result of the build before doing a new patch set | 12:33 |
| mattt | here we go | 12:35 |
| odyssey4me | mattt in my test AIO it appears to be working | 12:36 |
| odyssey4me | yup, horizon play is done with no errors | 12:36 |
| mattt | ok great | 12:36 |
| mattt | so now back to master to see if the heat issue persists :P | 12:36 |
| * cloudnull happy dance | 12:36 | |
| odyssey4me | ok, so we remove the deps and see whether that passes now? | 12:37 |
| mattt | cool! | 12:37 |
| odyssey4me | oh, fail? | 12:38 |
| mattt | WUUUUT | 12:38 |
| mattt | these are debian packages | 12:39 |
| odyssey4me | failed: [aio1] => (item=curl,gcc,git-core,liberasurecode1,liberasurecode-dev,libffi-dev,openssh-server,python-dev,rsync) => {"attempts": 5, "failed": true, "item": "curl,gcc,git-core,liberasurecode1,liberasurecode-dev,libffi-dev,openssh-server,python-dev,rsync"} | 12:39 |
| * odyssey4me sighs | 12:40 | |
| odyssey4me | my aio passed, so it would seem this is a transitional error in openstack-ci | 12:40 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add release pinning to galera roles https://review.openstack.org/176720 | 12:41 |
| mattt | odyssey4me: i saw this error somewhere else today | 12:41 |
| mattt | so it's not the first time | 12:41 |
| cloudnull | miguelgrinberg's patch for heat testing has been running into this too | 12:41 |
| mattt | cloudnull: ah that's it | 12:41 |
| odyssey4me | updated review without the explicit deps | 12:41 |
| mattt | i can't tell what it's actually failing on tho | 12:43 |
| mattt | http://logs.openstack.org/20/176720/7/check/os-ansible-deployment-dsvm-check-commit/ca7cc54/logs/ansible-logging/ansible.log doesn't reveal much | 12:43 |
| odyssey4me | yeah, my AIO even passed tempest tests so it seems that we may have finally beaten down the kilo branch gate block | 12:44 |
| mattt | so master is failing on heat install due to | 12:57 |
| mattt | Could not find a version that satisfies the requirement python-saharaclient>=0.9.0 (from heat) (from versions: 0.8.0) | 12:57 |
| mattt | we only have 0.8.0 | 12:57 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Set gerrit default branch to kilo https://review.openstack.org/176246 | 12:57 |
| mattt | going to bump it in master | 12:58 |
| odyssey4me | stevelle are you prepping a patch for master for the horizon local_settings changes needed? | 13:00 |
| odyssey4me | I think we need to get that in and backport it to kilo before we do the next kilo rc | 13:00 |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Bump python-saharaclient https://review.openstack.org/177241 | 13:04 |
| mattt | failing to create launchpad bugs, so no Closes-Bug in the above review | 13:04 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Add package list and repos to AIO diagnostics https://review.openstack.org/176657 | 13:06 |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Bump python-saharaclient https://review.openstack.org/177241 | 13:08 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Removed hard coded package pinning for mariadb/galera https://review.openstack.org/177259 | 13:27 |
| mattt | bah, what is this swift failure | 13:33 |
| odyssey4me | mattt cloudnull yep: http://logs.openstack.org/20/176720/8/check/os-ansible-deployment-dsvm-check-commit/328bb37/console.html#_2015-04-24_13_28_33_821 | 13:37 |
| *** KLevenstein has joined #openstack-ansible | 13:37 | |
| odyssey4me | probably another upstream mirror bork | 13:37 |
| *** andymccr has joined #openstack-ansible | 13:42 | |
| odyssey4me | mattt cloudnull so the packages being installed at the time are: liberasurecode1,liberasurecode-dev | 13:45 |
| odyssey4me | the other packages in the list are already there | 13:46 |
| *** Mudpuppy has joined #openstack-ansible | 13:47 | |
| cloudnull | yea, curious if those are being missed due to a repo not enabled on the new dsvm image? | 13:47 |
| mattt | it seems to download them tho doesn't it | 13:47 |
| odyssey4me | yeah, it seems to actually do all the bits in the container - but it fails on the host with no output | 13:48 |
| odyssey4me | so cloudnull may be right - it's another funny in the image | 13:48 |
| mattt | see bottom of http://logs.openstack.org/20/176720/8/check/os-ansible-deployment-dsvm-check-commit/328bb37/logs/ansible-logging/ansible.log | 13:48 |
| odyssey4me | maybe missing the backports pool - it seems to be coming from there in the container | 13:48 |
| mattt | oh derp, that's the container :( | 13:49 |
| *** Mudpuppy has quit IRC | 13:49 | |
| mattt | so installs successfully in container but not on host | 13:49 |
| cloudnull | we could also re-enable the retry max for the successerator which will enable super debug on failure. | 13:49 |
| *** Mudpuppy has joined #openstack-ansible | 13:49 | |
| odyssey4me | this will tell us more about the repositories and packages: https://review.openstack.org/176657 | 13:50 |
| odyssey4me | we just have to wait for it to finish | 13:50 |
| mattt | nice | 13:51 |
| *** yaya has joined #openstack-ansible | 13:52 | |
| odyssey4me | confirmed that both packages happen to be from trusty-backports/universe and don't exist in any other pool | 13:52 |
| odyssey4me | odd: this one worked and has the other patch as a dependant: https://review.openstack.org/176246 | 13:53 |
| odyssey4me | rax-ord vs hpcloud-b1 | 13:54 |
| mattt | yep | 13:54 |
| odyssey4me | rax-rd worked, hpcloud didn't | 13:54 |
| cloudnull | https://review.openstack.org/#/c/176246/ <- kilo is happy | 13:54 |
| odyssey4me | yup, it appears that we have another hpcloud surprise | 13:54 |
| mattt | cloudnull: about time | 13:54 |
| cloudnull | yay for hpcloud | 13:54 |
| mattt | so do we ask infra to include that repo in their apt sources? | 13:55 |
| mattt | or do we manually add it? | 13:55 |
| odyssey4me | check the initial apt-get updates in both logs | 13:55 |
| mattt | the latter would feel janky | 13:55 |
| odyssey4me | hpcloud has no backports repo | 13:55 |
| odyssey4me | mattt for now to get things moving we should add it if it isn't there so we don't get more surprises | 13:55 |
| cloudnull | so we could, in the swift role, add the backports as its really only needed there | 13:55 |
| odyssey4me | but yeah we should let openstack-infra know | 13:55 |
| odyssey4me | cloudnull I'd suggest doing it in the AIO bootstrap | 13:56 |
| odyssey4me | otherwise we're going down the road of doing deployer stuff again in the ansible things | 13:56 |
| *** stevemar has joined #openstack-ansible | 13:57 | |
| cloudnull | this is true. we'll need to note something somewhere that swift, for kilo, needs the backports repo . | 13:57 |
| cloudnull | as this will likely be a problem in production if a host is not already setup to consume ubuntu backports | 13:57 |
| mattt | is there no way to pull in swift with all the erasurecoding stuff? | 13:57 |
| mattt | since we're not using it | 13:58 |
| mattt | *witout | 13:58 |
| odyssey4me | mattt yeah I suppose we could, but what happens if a customer does want to configure it? | 13:58 |
| odyssey4me | *deployer | 13:58 |
| mattt | andymccr: is that a facility that's enabled in teh swift config? | 13:58 |
| mattt | or does it just work out of the bo? | 13:58 |
| mattt | *box | 13:58 |
| mattt | dang can't type today | 13:58 |
| andymccr | i dont think it just works mattt and its not setup atm | 13:59 |
| andymccr | cos it was only just added as a feature | 13:59 |
| andymccr | but tbh i havnt looked at it so im not 100% sure | 13:59 |
| andymccr | id be surprised if it just works out thebox though | 13:59 |
| odyssey4me | is anyone logging a bug for this already, or shall I? | 13:59 |
| andymccr | and atm you cant configure the ring to use erasurecode | 13:59 |
| cloudnull | mattt also its a py dep https://github.com/openstack/swift/blob/master/requirements.txt#L12 which has a hard dependency on the library being available https://bitbucket.org/kmgreen2/pyeclib/src/041d44311bf3d7d3f655dfc6c692d8693b9e0634/setup.py?at=master#cl-62 | 13:59 |
| mattt | that sucks cloudnull | 13:59 |
| mattt | odyssey4me: i can log a bug | 13:59 |
| mattt | andymccr: ok but guess we have to pull it in :( | 13:59 |
| odyssey4me | shall I prep a patch for the AIO bootstrap? | 14:00 |
| cloudnull | +1 | 14:00 |
| andymccr | yeh mattt i think its a swift dep itself isnt it? | 14:00 |
| mattt | odyssey4me: https://bugs.launchpad.net/openstack-ansible/+bug/1448152 | 14:02 |
| openstack | Launchpad bug 1448152 in openstack-ansible "swift installation failing in kilo branch" [Undecided,New] | 14:02 |
| *** sigmavirus24_awa is now known as sigmavirus24 | 14:03 | |
| mattt | this is presumably an issue on master too i'd imagine | 14:03 |
| cloudnull | +1 | 14:04 |
| andymccr | and not just for us i'd imagine? | 14:05 |
| Sam-I-Am | kind of shitty when places have different images | 14:05 |
| odyssey4me | any thoughts on the best way to figure out of the backports repo is active - I'm thinking a grep of the output of 'apt-get -q update' at this stage because a trawl of the config files is less reliable | 14:09 |
| mattt | odyssey4me: if a grep doesn't find it in /etc/apt/sources* then add it | 14:09 |
| andymccr | cant we just ensure it is? | 14:09 |
| andymccr | it shouldnt then add it twice | 14:10 |
| cloudnull | odyssey4me: # if ! grep -rni backports /etc/apt/sources.list*; then echo blah; fi | 14:11 |
| odyssey4me | cloudnull yeah, but the backports word could be in text description or in a commented line | 14:12 |
| odyssey4me | perhaps this is ok? grep -rni "^deb.*backports" /etc/apt/sources.list* | 14:13 |
| cloudnull | odyssey4me: # if ! grep -rn '^deb' /etc/apt/sources.list*| grep -i backports; then echo blah; fi | 14:13 |
| andymccr | this is just for the gate? | 14:13 |
| odyssey4me | andymccr yep | 14:13 |
| andymccr | why cant we add an ansible task that just ensures that repo is there? | 14:13 |
| andymccr | like if its a common thing its not going to just be a thing for the gate | 14:13 |
| odyssey4me | andymccr because then we're doing deployer work | 14:14 |
| odyssey4me | we're interfering with the ubuntu repository configuration which for everything else we assume is already in place | 14:14 |
| andymccr | hmm | 14:14 |
| andymccr | but as an example we have a default for mariadb repo - which isnt included by default in the ubuntu repo, you can set this as the deployer so its not hardcoded | 14:15 |
| andymccr | the same would apply if you were trying to deploy the erasurecode packages which are not by default in there unless you include -backports | 14:15 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add ubuntu backports repository to AIO if it's not present https://review.openstack.org/177284 | 14:24 |
| odyssey4me | andymccr but the mariadb repo is something we added, the ubuntu is not a repo we added | 14:25 |
| andymccr | odyssey4me: sure, but the principle is the same - if the default is not always going to ensure that erasurecode packages are there - it would be helpful for the deployer to be able to configure the repository that contains those | 14:25 |
| andymccr | we can set a sensible default as -backports, and you can even choose not to add it if the var doesnt exist | 14:26 |
| odyssey4me | but then we're adding stuff to configure things that a deployer could do themselves and we step into the realm of fielding issues relating to some random thing they added | 14:26 |
| andymccr | how? i dont see how thats any different to a deployer changing the mariadb repo and then asking us to field an issue on something random they added | 14:27 |
| odyssey4me | we can't really set a sensible default either, I don't think - the patch I've done ^ grabs the info from the existing environment instead of relying on some arbitrary default set in the project | 14:27 |
| andymccr | put it this way if somebody tries to deploy os-a-d today and runs into that issue, they should manually add the repo? | 14:27 |
| odyssey4me | yes, they should have a play/role as part of their general infrastructure management which handles repository configuration | 14:28 |
| andymccr | thats the first time we've needed that though if im not mistaken? | 14:28 |
| odyssey4me | something that helps them do apt-get upgrades and all that | 14:28 |
| odyssey4me | andymccr yeah, this is thanks to the recent refresh of images in openstack-ci | 14:28 |
| andymccr | its also because the only packages from the backports repo that we needed are for the erasurecode stuff. | 14:29 |
| andymccr | and they got added in the latest swift release | 14:29 |
| andymccr | so before that you wouldnt have seen those issues | 14:29 |
| mattt | odyssey4me: apt-get --print-uris update | 14:29 |
| odyssey4me | mattt ooh, that's pretty cool | 14:30 |
| odyssey4me | do you think that'd be preferred over grep -qrni "^deb.*backports" /etc/apt/sources.list* ? | 14:31 |
| odyssey4me | I need to update the kilo patch with this, so I'd like to get it right | 14:32 |
| mattt | odyssey4me: i think either is fine, but do "^deb .*" | 14:32 |
| mattt | so you're not matching on deb-src | 14:32 |
| odyssey4me | andymccr fyi if you add a second copy of the repo source you get 'Duplicate sources.list entry' errors for apt | 14:33 |
| mattt | "^deb .*backports" | 14:33 |
| andymccr | id hope the ansible module doesnt let you add it twice? | 14:33 |
| odyssey4me | mattt good call | 14:33 |
| mattt | odyssey4me: may be worth being explicit and saying trusty-backports | 14:34 |
| andymccr | it just seems weird to me that we're adding it to containers but not hosts. | 14:34 |
| odyssey4me | mattt good call | 14:34 |
| andymccr | so we're saying containers need the -backports but hosts arent getting it cos oyu shouldve added it manually? | 14:34 |
| mattt | andymccr: problem is the host is preconfigured, the container isn't | 14:35 |
| mattt | it would be complicated to _update_ someone's apt sources | 14:35 |
| mattt | and you certainly can't blow it away and recreate | 14:35 |
| andymccr | mattt: theres an apt module isnt there? | 14:36 |
| mattt | andymccr: open source patches welcome? | 14:36 |
| mattt | :) | 14:36 |
| andymccr | im trying to figure out why nobody else thinks thats a sensible idea? | 14:36 |
| andymccr | im clearly missing something | 14:36 |
| *** galstrom_zzz is now known as galstrom | 14:36 | |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add release pinning to galera roles https://review.openstack.org/176720 | 14:38 |
| odyssey4me | andymccr I think it's a great idea for something outside OSAD like rpc-extras, but not for OSAD | 14:38 |
| andymccr | odyssey4me: ok so why? like if we're defining it for the containers | 14:38 |
| andymccr | which we are. | 14:38 |
| odyssey4me | I don't even think we should be as prescriptive as we are for the containers | 14:38 |
| odyssey4me | the repo used is a deployer's choice | 14:39 |
| odyssey4me | in a previous life we refused to use backport repositories in order to reduce risk and increase stability - it was seen by some that backports weren't adequately tested | 14:40 |
| andymccr | we're not letting that happen for the containers though? aside from that you could change a var, which is great - but you could use that same var | 14:40 |
| mattt | andymccr: oh you meant ansible module, not python module | 14:40 |
| andymccr | mattt: yeah | 14:40 |
| odyssey4me | my mission right now is to unblock the gate though - if it's decided to do it differently that's great, but the decision needn't be made right now | 14:40 |
| mattt | andymccr: yeah you could easily do this w/ ansible, but it's likely that the host's sources aren't configured with the same apt sources that are defined in os-a-d by default | 14:41 |
| palendae | The gate is blocked because we bumped the swift SHAs again right? | 14:41 |
| andymccr | my point really is that as it stands it wont deploy at all without the backports repo - so anybody trying to adopt os-a-d may run into this issue and just be like f-it | 14:41 |
| mattt | andymccr: so do you just blindly add a backports repo if it's missing, or do you figure out what other ubuntu mirror is being used, then use that? | 14:41 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add release pinning to galera roles https://review.openstack.org/176720 | 14:42 |
| andymccr | mattt: we're setting one for the containers, id just use that one | 14:42 |
| odyssey4me | palendae nope, the gate is blocked because of loads of changes upstream in various ways | 14:42 |
| andymccr | it seems odd that somebody would want a diff repo for containers and hosts, so if htey havnt set it, they probably dont care? | 14:42 |
| mattt | andymccr: yeah we could, i'm not aware of any glaringly obvious reasons why that'd be bad | 14:42 |
| palendae | odyssey4me: Because we bumped SHAs... | 14:42 |
| odyssey4me | palendae that may be, but for a kilo RC we need to bump sha's regardless and deal with the consequences | 14:43 |
| mattt | andymccr: we should get more input then | 14:43 |
| mattt | andymccr: and if the consensus is to add it then we get a review in | 14:44 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add ubuntu backports repository to AIO if it's not present https://review.openstack.org/177284 | 14:45 |
| palendae | I came in halfway through this discussion, but are you guys suggesting that repos be left up to the deployer? If so, would OSAD even work standalone at that point? | 14:45 |
| mattt | andymccr: what if we bail if the repo doesn't exist? | 14:46 |
| odyssey4me | palendae my suggestion is that ubuntu repositories not be adjusted in our playbooks/roles | 14:46 |
| mattt | w/ a helpful message for the deployer | 14:46 |
| mattt | andymccr: is that a reasonable compromise ? | 14:46 |
| andymccr | mattt: naw cos if you had controlled your repos in a clever way and had those packages in a repo that wasnt "backports" it'd fail unnecessarily | 14:46 |
| lbragstad | cloudnull: quick ansible question for you when you have a minute. The current keystone-deploy stuff is broken up into roles (sql/http/memcache/etc..) but we also want the ability to support Centos. Do you have a recommendation for organizing the plays without doubling each role? | 14:46 |
| palendae | odyssey4me: Ok, so for libraries like erasure coding that are necessary for swift, how do we get them? | 14:47 |
| lbragstad | cc dolphm ^ | 14:47 |
| odyssey4me | mattt then we get stuck in trying to keep up an exhaustive checklist in our plays, which is a very bad idea IMO | 14:47 |
| dstanek | lbragstad: i do it all of the time by keeping the dependent packages separately | 14:48 |
| odyssey4me | lbragstad why not just add the code to the same roles, but adjust the tasks to accommodate the differences between platforms? | 14:48 |
| odyssey4me | if execute task when platform = ubuntu | 14:48 |
| odyssey4me | *ie | 14:48 |
| cloudnull | lbragstad to add additional OS support you'd need to add in seperation logic into the various roles and a new package set. | 14:48 |
| lbragstad | dstanek: you're recommending the when ansible_os_family == 'Debian' path, right/ | 14:48 |
| cloudnull | ^ yes | 14:48 |
| odyssey4me | palendae we fail if the deployer's apt repository doesn't have the packages we want - just like we would for any other packages | 14:49 |
| andymccr | odyssey4me: yeah but we have a working by default thing (before this) | 14:49 |
| lbragstad | cloudnull: odyssey4me dstanek cool, that's the path that I start working down but it doubled the playbook sizes. I just wasn't sure if there was a better way? | 14:49 |
| lbragstad | http://cdn.pasteraw.com/7zp564z3pqylyj7pniaovnyiuln32hz | 14:49 |
| palendae | odyssey4me: But for OSAD | 14:49 |
| andymccr | so does the default ubuntu image for 14.04 include the -backports repo by default? | 14:49 |
| palendae | Are we defining nothing? | 14:49 |
| odyssey4me | best we can do is note that the backports repo is required on swift containers and hosts | 14:49 |
| odyssey4me | andymccr yes | 14:50 |
| andymccr | it does? then i concur with your patch | 14:50 |
| palendae | Or are we saying OSAD is no longer installable by itself? | 14:50 |
| odyssey4me | yup, the default Ubuntu deployment includes the backports repo - if it's not there then someone has changed it | 14:50 |
| andymccr | if its already there on a default ubuntu 14.04 install then i think its acceptable | 14:50 |
| dstanek | lbragstad: cloudnull: i do stuff like this: http://paste.openstack.org/show/205716/ | 14:50 |
| andymccr | which image for gating is failing? | 14:50 |
| mattt | andymccr: HP pub cloud image | 14:51 |
| andymccr | always :/ | 14:51 |
| andymccr | ok that makes sense at least | 14:51 |
| odyssey4me | yup, no surprises there :p | 14:51 |
| lbragstad | dstanek: ah... that makes sense | 14:51 |
| dolphm | dstanek: how are you using the 'packager' variable, for example? | 14:52 |
| cloudnull | dstanek you can break them up into vars, but you'd still need logic in the tasks to go to different tasks for rhel vs ubuntu. | 14:52 |
| dolphm | dstanek: using it in a 'when' for an apt task? | 14:52 |
| dstanek | forgot the important part: http://paste.openstack.org/show/205717/ | 14:52 |
| lbragstad | {{ packager }} would be defined at the http role layer? | 14:52 |
| dolphm | cloudnull: breaking things up by distro sounds like a slippery slope to non-dry-ness | 14:52 |
| cloudnull | it is | 14:53 |
| lbragstad | :-/ | 14:53 |
| dolphm | well then it sounds like an anti-pattern :P | 14:53 |
| cloudnull | but it would really only be for the instalaltion of OS packages. | 14:53 |
| dstanek | cloudnull: i have not come across the need to do that yet - my ansible needs so far have been to setup my dev environments and not control an infrastructure :-) | 14:53 |
| dstanek | putting the packages and packages in vars should be enough in this case | 14:54 |
| dstanek | s/packages and packages/packager and packages/ | 14:54 |
| cloudnull | yes. but the task to install said packages, ubuntu vs centos, would be different and need something like a conditional when ansible_os_family == 'Debian' | 14:55 |
| dolphm | dstanek: but there are more differences between platforms than just package names :( | 14:55 |
| cloudnull | etc.. | 14:55 |
| dstanek | dolphm: i've only found a few that i care about - http service name, a handful of paths, etc. | 14:55 |
| cloudnull | dolphm there are also init scripts, and a few others, but the bulk of our current roles should work. | 14:55 |
| lbragstad | so, the other part I was thinking about was being able to isolate distro specific plays in a distro specific role? | 14:55 |
| dstanek | all configuration stuff - so my playbooks are still dry | 14:56 |
| dolphm | lbragstad: let's keep both OS's all in one task file until it works, and then look to see if it makes sense to break it apart after | 14:56 |
| lbragstad | dolphm: ++ I agree with that | 14:56 |
| palendae | Is there a spec for adding CentOS support to OSAD? | 14:57 |
| sigmavirus24 | so | 14:57 |
| sigmavirus24 | I would suggest the first spec should be to refactor OSAD to allow for distro-specific roles | 14:57 |
| dolphm | palendae: no, this isn't OSAD | 14:58 |
| sigmavirus24 | Then to add specs for specific distros | 14:58 |
| dolphm | palendae: looking to deploy keystone to centos for public cloud | 14:58 |
| dolphm | sigmavirus24: ^ | 14:58 |
| sigmavirus24 | roger that | 14:58 |
| lbragstad | sigmavirus24: palendae based on the exiting keystone-deploy work which targets debian (we test with Travis) https://github.com/dolph/keystone-deploy | 14:59 |
| palendae | Gotcha | 14:59 |
| cloudnull | lbragstad dolphm dstanek Im quite fond of single role with conditional logic in the main tasks. | 14:59 |
| cloudnull | IE https://github.com/stackforge/os-ansible-deployment/blob/master/playbooks/roles/os_nova/tasks/main.yml | 14:59 |
| cloudnull | another example https://github.com/stackforge/os-ansible-deployment/blob/master/playbooks/roles/os_nova/tasks/nova_compute.yml | 15:00 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add ubuntu backports repository to AIO if it's not present https://review.openstack.org/177284 | 15:00 |
| dolphm | cloudnull: ack | 15:00 |
| cloudnull | in that case centos vs ubuntu would == kvm vs qemu | 15:00 |
| mancdaz | cloudnull not that I'm saying it's right or wrong, but that kind of logic was what people didn't like about rcbops chef cookbooks | 15:01 |
| dolphm | those are both pretty elegant | 15:01 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add release pinning to galera roles https://review.openstack.org/176720 | 15:01 |
| cloudnull | mancdaz i'd argue that the osops_utils usage ontop of that kind of logic is what people didnt like about the chef cookbooks. | 15:01 |
| palendae | mancdaz: I haven't seen a solution that wasn't kinda dirty in any of the config management things | 15:01 |
| dstanek | cloudnull: i like having the playbooks be the recipe template and keeping the configuration is vars files | 15:02 |
| palendae | I didn't like how puppet did it either | 15:02 |
| mancdaz | cloudnull ok I should have said 'one of the things' :P | 15:02 |
| cloudnull | :-) | 15:02 |
| mancdaz | a lot of the discussion was based around the fact that you should know what a role is going to do | 15:03 |
| sigmavirus24 | mancdaz: since when were there things people didn't like about the chef cookbooks? | 15:03 |
| mancdaz | if you want it to do something else, you have a different role | 15:03 |
| openstackgerrit | Jesse Pretorius proposed stackforge/os-ansible-deployment: Add release pinning to galera roles https://review.openstack.org/176720 | 15:03 |
| mancdaz | actually role/role is not interchangeable between chef and ansible, but for the purposes of this | 15:03 |
| cloudnull | mancdaz this is true, but that would mean that roles would be broken up into os_nova_ubuntu os_nova_rhel etc.. | 15:04 |
| mancdaz | well ok I missed the bit where this was about different oses | 15:04 |
| lbragstad | dolphm: are you leaning on way or the other? | 15:04 |
| dstanek | cloudnull: are things that different so that the vars trick won't work? | 15:04 |
| mancdaz | I was looking at the nova example | 15:04 |
| sigmavirus24 | cloudnull: would it be possible to do a conditional include based on the operating system inside the os_nova role? | 15:05 |
| dolphm | lbragstad: i think so - cram everything into one task file, then look at the platform differences and see how it makes sense to refactor into something like https://github.com/stackforge/os-ansible-deployment/blob/master/playbooks/roles/os_nova/tasks/main.yml afterword | 15:05 |
| cloudnull | dstanek the task name are different between rhel and deb | 15:05 |
| cloudnull | sigmavirus24 yes | 15:05 |
| mancdaz | in chef we used a platform hash that contained all the things | 15:05 |
| cloudnull | IE what packages are installed and how init systems work . | 15:05 |
| lbragstad | dolphm: cool, well in that case, I'll keep hacking on http://cdn.pasteraw.com/7zp564z3pqylyj7pniaovnyiuln32hz | 15:05 |
| mancdaz | and then just populated that inside the cookbook depending on what platform you were running - is this approach not possible in ansible? | 15:06 |
| dstanek | mancdaz: that's the approach i was advocating | 15:06 |
| lbragstad | another question I had, after we have one big playbook with tasks for both distros, how would we go about adding support for nginx versus apache and uwsgi versus mod_wsgi? | 15:06 |
| galstrom | we could use a case statement on the os platform.. then use a packages hash with top level keys of the os platforms.. | 15:06 |
| dstanek | mancdaz: see my two links from above | 15:06 |
| cloudnull | lbragstad what you can do is the conditional on the os for a particular included file. which would limit the use of things like "when: ansible_os_family == "Debian"" | 15:06 |
| galstrom | that lets you map package name differences between os's | 15:07 |
| mancdaz | dstanek ah ok yeah | 15:07 |
| cloudnull | lbragstad ie https://github.com/stackforge/os-ansible-deployment/blob/master/playbooks/roles/os_nova/tasks/nova_compute.yml#L16-L17 | 15:07 |
| cloudnull | then do all of the deb things in that one include or all of the rhel things in some other include. | 15:07 |
| dolphm | cloudnull: ++ we'll work towards that | 15:08 |
| cloudnull | galstrom you stop your chef sillyness. :) | 15:09 |
| galstrom | cloudnull :) | 15:09 |
| galstrom | that is how it is done sir | 15:09 |
| cloudnull | distro names in a hash and conditional logic wrapped up in a syscall to ruby == winning | 15:10 |
| *** sandywalsh has quit IRC | 15:10 | |
| cloudnull | put a little osops_utils on top and an upstream mysql cookbook / gem and you have yourself a recipe for a good time. | 15:11 |
| *** jwagner_away is now known as jwagner | 15:12 | |
| andymccr | who doesnt love a good time | 15:13 |
| palendae | chef's all about recipes for a good time | 15:13 |
| palendae | Why'd you guys get rid of good times? | 15:13 |
| dstanek | cloudnull: so in my case instead of using vars for the configuration i would move that logic into the task? | 15:14 |
| *** yaya has quit IRC | 15:15 | |
| cloudnull | dstanek in your case you could use both, include the vars based on the OS and then also conditionally do things based on the platform family. you'd have a common set of tasks that could be done on both platforms and conditional tasks. | 15:17 |
| openstackgerrit | Miguel Alejandro Cantu proposed stackforge/os-ansible-deployment-specs: Add spec for BP implement-ceilometer https://review.openstack.org/169417 | 15:20 |
| *** yaya has joined #openstack-ansible | 15:26 | |
| *** jlvillal has quit IRC | 15:31 | |
| *** sacharya has joined #openstack-ansible | 15:33 | |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Updates for tag 10.1.4 https://review.openstack.org/177050 | 15:34 |
| openstackgerrit | Jacob Wagner proposed stackforge/os-ansible-deployment-specs: added spec build-facts-archive https://review.openstack.org/177023 | 15:35 |
| *** galstrom is now known as galstrom_zzz | 15:39 | |
| *** galstrom_zzz is now known as galstrom | 15:40 | |
| *** KLevenstein has quit IRC | 15:45 | |
| *** sdake has joined #openstack-ansible | 15:51 | |
| *** KLevenstein has joined #openstack-ansible | 15:52 | |
| mattt | odyssey4me: lol, the swift fix just failed on a heat issue | 15:53 |
| sigmavirus24 | inconceivable | 15:53 |
| odyssey4me | mattt bah, have you guys figured out the heat issue - is it transient? | 15:54 |
| odyssey4me | another dependency failure? | 15:55 |
| mattt | odyssey4me: i put in https://review.openstack.org/#/c/177241/, but it failed on the swift stuff | 15:56 |
| mattt | we probably need to bundle these together | 15:56 |
| mattt | i thought your review was targetting kilo specifically | 15:57 |
| mattt | odyssey4me: i'm going to update https://review.openstack.org/#/c/177284/ to bump saharaclient | 15:57 |
| odyssey4me | mattt go for it | 15:57 |
| odyssey4me | we have a successful build check for https://review.openstack.org/176720 mattt andymccr cloudnull palendae hughsaunders d34dh0r53 can we get some reviews please to unblock the gate for kilo, finally? | 16:01 |
| odyssey4me | it's still copying logs or something in zuul, in case you're wondering where the result logs are | 16:01 |
| hughsaunders | ahh | 16:01 |
| alextrcitiy | Hey, is master gate still blocked? | 16:02 |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Add ubuntu backports repository to AIO if it's not present https://review.openstack.org/177284 | 16:02 |
| mattt | alextrcitiy: yessir | 16:02 |
| alextrcitiy | Ah. thanks | 16:02 |
| odyssey4me | alextrcitiy this patch is aiming to unblock master: https://review.openstack.org/177284 | 16:04 |
| alextrcitiy | Cool. I'll keep a close eye on it | 16:04 |
| mattt | why do we keep rechecking things like https://review.openstack.org/176166 ? | 16:05 |
| mattt | :) | 16:05 |
| odyssey4me | hmm, it looks like the zuul status page is broken | 16:05 |
| odyssey4me | haha mattt good question, I guess it's a habit :p | 16:06 |
| *** galstrom is now known as galstrom_zzz | 16:06 | |
| mattt | afk for a bit | 16:06 |
| andymccr | with that fix, do we have to bump the sha's and therefore have to do the swift fix at the same time? | 16:09 |
| andymccr | nm its already been approved | 16:09 |
| odyssey4me | andymccr master tracks master, so no sha's | 16:09 |
| andymccr | this is kilo though | 16:10 |
| sigmavirus24 | mattt: we keep rechecking it because we're optimists | 16:10 |
| odyssey4me | andymccr oh, that one - part of the reason for bumping the sha's is that the build wasn't working due to the requirements changes - so we can't pin mariadb on its own | 16:11 |
| andymccr | ok - i mean ideally i would prefer the swift fix and the galera fix to be separate patches but if its not possible its not possible | 16:11 |
| *** Mudpuppy has quit IRC | 16:12 | |
| odyssey4me | I suppose that technically we could have not bumped the sha's for everything and only done the global requirements, but that would've meant another round of testing | 16:12 |
| *** sandywalsh has joined #openstack-ansible | 16:13 | |
| odyssey4me | yeah, if we had caught the galera issue sooner and got the patch in before horizon's libraries upped their versions, we would have had an easier time | 16:13 |
| andymccr | well its done now so its purely academic discussing it. but 1:1 issue:PR ratio is how i would personally prefer it. | 16:13 |
| odyssey4me | andymccr yup, agreed :) | 16:13 |
| odyssey4me | we need to try and respond more quickly next time | 16:13 |
| sigmavirus24 | andymccr: I think that's how everyone would prefer it but in reality it isn't always possible =( | 16:14 |
| andymccr | sigmavirus24: i agree which is why im asking if it was possible - it felt like it might've been there | 16:14 |
| andymccr | but that was before i saw it was approved already :P | 16:15 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Update for tag 9.0.8 https://review.openstack.org/177051 | 16:16 |
| sigmavirus24 | We should build a system that otter-approves patches | 16:16 |
| hughsaunders | otter? | 16:16 |
| odyssey4me | otrs? | 16:17 |
| *** Mudpuppy has joined #openstack-ansible | 16:17 | |
| odyssey4me | haha cloudnull nice sneak-tweak :p | 16:17 |
| sigmavirus24 | hughsaunders: bad reference to https://github.com/rackerlabs/otter | 16:17 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Removed hard coded package pinning for mariadb/galera https://review.openstack.org/177259 | 16:18 |
| hughsaunders | okays | 16:18 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Update for tag 9.0.8 https://review.openstack.org/177051 | 16:18 |
| sigmavirus24 | hughsaunders: you don't have to like the joke, you just have to laugh | 16:18 |
| cloudnull | there we go. | 16:18 |
| cloudnull | i dep'd one on the other. | 16:18 |
| cloudnull | icehouse has convergence issues and i think we can safely affirm that the ssh retry is not enough. | 16:20 |
| *** javeriak has joined #openstack-ansible | 16:20 | |
| odyssey4me | cloudnull yep, for icehouse especially | 16:21 |
| odyssey4me | we don't seem to have the same reliability issues for kilo | 16:21 |
| *** javeriak has quit IRC | 16:21 | |
| cloudnull | nope | 16:21 |
| *** daneyon_ has joined #openstack-ansible | 16:22 | |
| *** galstrom_zzz is now known as galstrom | 16:22 | |
| *** KLevenstein has quit IRC | 16:24 | |
| *** daneyon has quit IRC | 16:25 | |
| hughsaunders | "fatal: [node07_rsyslog_container-55585dda] => SSH Error: data could not be sent to the remote host. Make sure this host can be reached over ssh" QQ | 16:33 |
| *** javeriak has joined #openstack-ansible | 16:33 | |
| * hughsaunders contemplates putting ssh_retry into jenkins-rpc | 16:33 | |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Add basic orchestration tests to the gate check https://review.openstack.org/176166 | 16:36 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177160 | 16:36 |
| openstackgerrit | Miguel Alejandro Cantu proposed stackforge/os-ansible-deployment: Implement Ceilometer[WIP] https://review.openstack.org/173067 | 16:36 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Make script executable and BSD sed compatible https://review.openstack.org/177115 | 16:36 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Add package list and repos to AIO diagnostics https://review.openstack.org/176657 | 16:37 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Set external roles to import into ansible namespace https://review.openstack.org/177131 | 16:37 |
| cloudnull | ^ updated dep to ensure the repo/req bump is in place. | 16:39 |
| cloudnull | hughsaunders thats not a bad idea for the jenkins-rpc sy stem | 16:40 |
| cloudnull | *system | 16:40 |
| cloudnull | sigmavirus24 andymccr https://review.openstack.org/#/q/owner:%22Ian+Cordasco%22+status:open+project:stackforge/os-ansible-deployment,n,z <- do you guys have a consensus on the glance patches ? | 16:42 |
| sigmavirus24 | other than the ones failing? | 16:42 |
| sigmavirus24 | oh yeah, I need to update those | 16:42 |
| sigmavirus24 | sorry, got side tracked yesterday | 16:42 |
| * sigmavirus24 heads off to fix the glance patches | 16:43 | |
| cloudnull | np. i didn't add the dep there because those patches were -1 on cr | 16:43 |
| andymccr | cloudnull: i think the sha bump can be abandoned, its already been applied in a separate patch | 16:43 |
| cloudnull | +1 | 16:43 |
| andymccr | and the others are a good solution to the problem so once thats fixed up its good to go. | 16:43 |
| andymccr | or at least a better solution to the problem (in my opinion) | 16:43 |
| cloudnull | okiedokie. | 16:47 |
| cloudnull | thanks andymccr sigmavirus24 | 16:47 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Set gerrit default branch to kilo https://review.openstack.org/176246 | 16:49 |
| sigmavirus24 | andymccr: before you hop off, do we want to stagger the times the cron job runs or should we just do */6? | 16:51 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177172 | 16:51 |
| * odyssey4me feeds zuul some more red bull and a jalapeno for a little extra fire | 16:51 | |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Add basic orchestration tests to the gate check https://review.openstack.org/177021 | 16:52 |
| andymccr | sigmavirus24: hmm... | 16:52 |
| sigmavirus24 | because */6 means every 6 hours | 16:52 |
| andymccr | yeh | 16:52 |
| sigmavirus24 | which is what we really wanted | 16:52 |
| sigmavirus24 | and changing the granularity is simpler that way than generating the comma-separated list | 16:52 |
| andymccr | does it matter if the image cache pruner runs at the same time on each host? | 16:52 |
| andymccr | youre gonna get round robin balanced to those nodes anyway | 16:52 |
| sigmavirus24 | andymccr: I'm not sure to be honest | 16:53 |
| *** sdake has quit IRC | 16:53 | |
| odyssey4me | andymccr sigmavirus24 maybe use ansible to randomise the minute in the cron job? | 16:53 |
| odyssey4me | so something like running it every 5 hours and <random> minutes | 16:54 |
| andymccr | imo simpler is better and i cant think of a reason why you'd need to have them run at diff times per host. | 16:54 |
| andymccr | sure that'd work | 16:54 |
| sigmavirus24 | sounds good to me | 16:54 |
| odyssey4me | maybe the random should be between 40 and 60 or something like that | 16:54 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Add release pinning to galera roles https://review.openstack.org/176720 | 16:56 |
| cloudnull | yay! kilo is now unblocked. | 16:58 |
| openstackgerrit | Ian Cordasco proposed stackforge/os-ansible-deployment: Run the glance cache cron jobs more frequently https://review.openstack.org/176368 | 16:59 |
| cloudnull | can we make this go https://review.openstack.org/#/c/177163/ | 17:00 |
| cloudnull | ^ odyssey4me d34dh0r53 palendae hughsaunders | 17:00 |
| openstackgerrit | Ian Cordasco proposed stackforge/os-ansible-deployment: Run the glance cache cron jobs more frequently https://review.openstack.org/176370 | 17:00 |
| sigmavirus24 | odyssey4me: didn't realize that you suggested between 40 and 60 | 17:01 |
| sigmavirus24 | right now it's going between 0 and 59 I believe | 17:01 |
| openstackgerrit | Ian Cordasco proposed stackforge/os-ansible-deployment: Run the glance cache cron jobs more frequently https://review.openstack.org/176365 | 17:04 |
| *** yaya has quit IRC | 17:05 | |
| odyssey4me | cloudnull I take it that https://review.openstack.org/177163 is essential in some way - how tested is it? | 17:09 |
| *** britthouser has joined #openstack-ansible | 17:10 | |
| cloudnull | i cant say that its been really put through its paces in overall testing but as it relates to the bug raised by rax support it should limit issues in contract tables . | 17:11 |
| cloudnull | from https://bugs.launchpad.net/openstack-ansible/+bug/1441363 | 17:11 |
| openstack | Launchpad bug 1441363 in openstack-ansible trunk "nf_conntrack schould be unloaded on swift object server" [Medium,In progress] - Assigned to Andy McCrae (andrew-mccrae) | 17:11 |
| cloudnull | Apsu if you could comment on the review https://review.openstack.org/177163 it'd be much appreciated. | 17:11 |
| odyssey4me | cloudnull actually I remember implementing this some time ago - at least I think it was this. | 17:12 |
| *** sdake has joined #openstack-ansible | 17:12 | |
| *** britthou_ has quit IRC | 17:12 | |
| odyssey4me | we still have master failures: https://review.openstack.org/177284 | 17:14 |
| odyssey4me | bah ssh failure | 17:14 |
| cloudnull | we could enable successerator for 1 retry ? but its kinda inevitable until we land ssh_retry upstream | 17:17 |
| odyssey4me | cloudnull it's so very rare though, and doing the retries more often ends up in masking issues | 17:17 |
| cloudnull | it does | 17:17 |
| cloudnull | hughsaunders whats the state of that upstream patch ? any movement ? | 17:18 |
| odyssey4me | happy for that for icehouse/juno because they're relatively mature - but for kilo/master I'm definitely against | 17:18 |
| cloudnull | ++ for sure. | 17:18 |
| odyssey4me | cloudnull hughsaunders it may be prudent to try and submit a similar PR for ansible 2 as well as 1.9; that way it may merge in 2 sooner :) | 17:18 |
| cloudnull | +1 | 17:19 |
| odyssey4me | this is assuming they have differing branches or something | 17:19 |
| odyssey4me | cloudnull it's a bit disconcerting to see the failures from the external CI | 17:22 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Implement Ceilometer[WIP] https://review.openstack.org/173067 | 17:22 |
| cloudnull | odyssey4me failure in # /usr/local/bin/pip install heat | 17:24 |
| cloudnull | which is one of our fixes. | 17:24 |
| odyssey4me | cloudnull on which one? | 17:24 |
| cloudnull | http://jenkins.propter.net/job/commit-multinode/648/consoleText | 17:24 |
| odyssey4me | cloudnull ah, but the multinode uses the rpc-mirror as I recall | 17:24 |
| odyssey4me | don't you need to rebuild the master repo there with the update? | 17:25 |
| cloudnull | yes i will once https://review.openstack.org/#/c/177284/4/playbooks/vars/repo_packages/openstack_clients.yml is in | 17:25 |
| odyssey4me | I might be wrong :p | 17:25 |
| cloudnull | it also dosent look like the external ci can process dependent changes. | 17:26 |
| cloudnull | hughsaunders? mattt ? | 17:26 |
| odyssey4me | cloudnull you may wish to check which patch set it's reporting on | 17:27 |
| odyssey4me | it doesn't cancel when a new patch set is uploaded | 17:27 |
| cloudnull | ah. maybe thats it | 17:27 |
| cloudnull | icehouse success https://review.openstack.org/#/c/177259/ <- typie typie make it go | 17:30 |
| odyssey4me | d34dh0r53 mattt andymccr https://review.openstack.org/177259 make the go! | 17:31 |
| andymccr | odyssey4me: say please | 17:32 |
| odyssey4me | hughsaunders palendae ^ | 17:32 |
| andymccr | ahahahaha | 17:32 |
| odyssey4me | andymccr with beer on top :p | 17:32 |
| andymccr | MANNERS SIR | 17:32 |
| odyssey4me | what manner of people are we, working on a friday night? | 17:32 |
| andymccr | the best. arrounnndd | 17:34 |
| cloudnull | andymccr :pretty please ?! | 17:34 |
| andymccr | cloudnull: i wanted it from odyssey4me :P | 17:34 |
| odyssey4me | cloudnull already done :p | 17:34 |
| cloudnull | oh, in that case GTY | 17:34 |
| cloudnull | :D | 17:34 |
| andymccr | hahaha | 17:34 |
| andymccr | https://www.youtube.com/watch?v=VIYqtkdMxQg | 17:35 |
| odyssey4me | now cloudnull will need to beg for https://review.openstack.org/177051 | 17:35 |
| andymccr | lgtm if it passes gate | 17:35 |
| cloudnull | andymccr +9000 | 17:36 |
| odyssey4me | andymccr yeah, it seems we'll see that shortly | 17:36 |
| cloudnull | your the best around! | 17:36 |
| andymccr | thats a song written about os-a-d | 17:37 |
| cloudnull | https://www.youtube.com/watch?v=oomCIXGzsR0 | 17:37 |
| odyssey4me | and that one's on the way too now :) | 17:37 |
| Apsu | cloudnull: Just got back from lunch. I'll take a look-see | 17:37 |
| Apsu | cloudnull: Done and doner. | 17:40 |
| cloudnull | tyvm! | 17:40 |
| Apsu | You bet | 17:41 |
| odyssey4me | sigmavirus24 have you finalised your glance cron patches yet? | 17:44 |
| stevelle | am I going to need to base any changes in master on https://review.openstack.org/#/c/177284/ (add backports repo) in order to get through the gate right now? | 17:58 |
| odyssey4me | stevelle yep | 18:00 |
| sigmavirus24 | odyssey4me: A) I thought you were leaving. B) No one concretely said to restrict the range from which the minute could be randomly selected | 18:00 |
| stevelle | I'll just hold that horizon change then | 18:01 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177163 | 18:02 |
| odyssey4me | sigmavirus24 I'm barely here, just waiting for https://review.openstack.org/177284 to gate | 18:02 |
| odyssey4me | alright, https://review.openstack.org/177284 is passed - just needs another +2+w to commit | 18:05 |
| odyssey4me | cloudnull anyone else still around? | 18:05 |
| odyssey4me | andymccr ^ ? | 18:05 |
| cloudnull | d34dh0r53 ^ | 18:05 |
| d34dh0r53 | on it | 18:06 |
| *** sandywalsh has quit IRC | 18:06 | |
| d34dh0r53 | Wherenow.org CI is now voting? | 18:07 |
| stevelle | yes | 18:07 |
| odyssey4me | d34dh0r53 only for code review, and it's only working properly on some branches | 18:07 |
| odyssey4me | it seems to be broken on master, and not responding to kilo | 18:07 |
| d34dh0r53 | ahh | 18:07 |
| cloudnull | and its only cr voting. | 18:08 |
| *** britthou_ has joined #openstack-ansible | 18:08 | |
| odyssey4me | it's set to vote +1/-1 whereas it previously was only set to vote +1 or nothing | 18:08 |
| *** sandywalsh has joined #openstack-ansible | 18:08 | |
| d34dh0r53 | cool, when did it change? | 18:08 |
| odyssey4me | d34dh0r53 during the course of today - around lunch time UK I think | 18:09 |
| d34dh0r53 | 👍 | 18:09 |
| *** britthouser has quit IRC | 18:09 | |
| *** erikmwilson is now known as Guest8853 | 18:11 | |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Run the glance cache cron jobs more frequently https://review.openstack.org/176365 | 18:11 |
| *** erikmwilson_ has joined #openstack-ansible | 18:11 | |
| cloudnull | sigmavirus24 updated dep so that the job will gate on https://review.openstack.org/#/c/176365/ | 18:12 |
| cloudnull | just so you know. | 18:12 |
| sigmavirus24 | Thanks | 18:12 |
| d34dh0r53 | Thanks to both of you | 18:12 |
| sigmavirus24 | opinions on whether the random selection should be from [0, 59] or [40, 59]? | 18:12 |
| d34dh0r53 | for being you | 18:12 |
| *** KLevenstein has joined #openstack-ansible | 18:12 | |
| sigmavirus24 | d34dh0r53: thank you for being a random location in memory =P | 18:12 |
| stevelle | [0, 59] | 18:13 |
| d34dh0r53 | sigmavirus24: I would actually do [1,59] as 0 will be a default for some jobs | 18:13 |
| odyssey4me | sigmavirus24 no strong opinion either way - I was basing the partial restriction based on your original set time | 18:13 |
| sigmavirus24 | d34dh0r53: good point. *starts condorcet poll to see if d34dh0r53's suggestion would win* | 18:14 |
| d34dh0r53 | lol | 18:14 |
| stevelle | fwiw [0,59] is the same as [0,14] for these purposes because you're going to run every 15 minutes, right? | 18:14 |
| stevelle | [1,59] just gives a slightly lower chance of not running a cycle on the hour | 18:14 |
| odyssey4me | stevelle every 6 hours as I recall | 18:15 |
| sigmavirus24 | stevelle: every 5 hours and a random number of minutes | 18:15 |
| stevelle | corrected: thats what I get for reading code at 3am | 18:15 |
| sigmavirus24 | so 5h random(1, 59) | 18:16 |
| sigmavirus24 | If no one objects to [1, 59], then I'm going to update the patches with that | 18:17 |
| odyssey4me | sigmavirus24 not me | 18:17 |
| odyssey4me | +1 | 18:17 |
| sigmavirus24 | And just to check my own Friday afternoon reading comprehension skills, http://docs.ansible.com/playbooks_filters.html#random-number-filter means that {{ 59 |random(start=1) }} would be what we want, correct? | 18:22 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Bump the tempest defcore repository version to master https://review.openstack.org/174983 | 18:23 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Removed hard coded package pinning for mariadb/galera https://review.openstack.org/177259 | 18:27 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Update for tag 9.0.8 https://review.openstack.org/177051 | 18:34 |
| mattt | FINALLY | 18:36 |
| mattt | FINALLY | 18:36 |
| *** jlvillal has joined #openstack-ansible | 18:42 | |
| sigmavirus24 | mattt: go home | 18:43 |
| cloudnull | peoples if your around, love pushing buttons, and reading code https://review.openstack.org/#/q/status:open+project:stackforge/os-ansible-deployment,n,z <- there are lots of master / kilo related reviews that need some reviewers :) | 18:48 |
| *** galstrom is now known as galstrom_zzz | 18:49 | |
| *** alextrcitiy has quit IRC | 18:53 | |
| openstackgerrit | Ian Cordasco proposed stackforge/os-ansible-deployment: Run the glance cache cron jobs more frequently https://review.openstack.org/176365 | 18:55 |
| *** erikmwilson_ is now known as erikmwilson | 18:55 | |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Add ubuntu backports repository to AIO if it's not present https://review.openstack.org/177284 | 18:57 |
| *** daneyon has joined #openstack-ansible | 18:58 | |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Add basic orchestration tests to the gate check https://review.openstack.org/176166 | 18:58 |
| *** daneyon_ has quit IRC | 18:59 | |
| openstackgerrit | Ian Cordasco proposed stackforge/os-ansible-deployment: Run the glance cache cron jobs more frequently https://review.openstack.org/176370 | 19:04 |
| d34dh0r53 | rabbit on gate seems to be failing a lot today | 19:05 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Updated icehouse for new dev work https://review.openstack.org/177384 | 19:06 |
| openstackgerrit | Ian Cordasco proposed stackforge/os-ansible-deployment: Run the glance cache cron jobs more frequently https://review.openstack.org/176368 | 19:09 |
| openstackgerrit | Steve Lewis proposed stackforge/os-ansible-deployment: Update Horizon configuration https://review.openstack.org/177386 | 19:13 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Updated icehouse for new dev work https://review.openstack.org/177384 | 19:15 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Updated juno for new dev work https://review.openstack.org/177388 | 19:15 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Updated juno for new dev work https://review.openstack.org/177388 | 19:17 |
| stevelle | I presume the fact that this task no longer exists in kilo/master is intentional https://github.com/stackforge/os-ansible-deployment/blob/juno/rpc_deployment/roles/container_archive/tasks/main.yml | 19:44 |
| stevelle | please correct me if I'm wrong | 19:44 |
| Sam-I-Am | cloudnull: ^ o master ptl | 19:46 |
| bgmccollum | Re MaaS rabbitmq_status.py, any plans to make CLUSTERED and CLUSTER_SIZE dynamic? | 19:55 |
| bgmccollum | its hard coded to True and 3 respectively. Which causes alerts for any environments that don't meet those criteria. | 19:55 |
| bgmccollum | bug already reported...thanks | 19:57 |
| openstackgerrit | Steve Lewis proposed stackforge/os-ansible-deployment: Update Horizon configuration https://review.openstack.org/177386 | 20:02 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177160 | 20:06 |
| *** nosleep77 has joined #openstack-ansible | 20:11 | |
| sigmavirus24 | stevelle: still trying to understand what relation https://bugs.launchpad.net/bugs/1439905 has to 177386 | 20:17 |
| openstack | Launchpad bug 1439905 in OpenStack Dashboard (Horizon) "[Launch Instance Fix] Show Admin Password" [Medium,In progress] - Assigned to Shaoquan Chen (sean-chen2) | 20:17 |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Add package list and repos to AIO diagnostics https://review.openstack.org/177434 | 20:17 |
| sigmavirus24 | mattt: stop working | 20:17 |
| stevelle | sigmavirus24: s/177386/177388/ ? | 20:19 |
| stevelle | err nm I pasted wrong place | 20:19 |
| stevelle | sigmavirus24: this is the basis of relation: https://github.com/openstack/horizon/commits/stable/kilo/openstack_dashboard/local/local_settings.py.example | 20:20 |
| sigmavirus24 | Launchpad is failing at showing that commit (given the only two that are linked are either a different bug or unmerged) | 20:23 |
| sigmavirus24 | Sorry for the hassle | 20:23 |
| stevelle | I see that the commit is not linked now. | 20:26 |
| stevelle | had overlooked that | 20:26 |
| sigmavirus24 | Yeah | 20:32 |
| sigmavirus24 | The reviews that are linked by jeepyb didn't seem related to local_settings.py.example which is why I was thoroughly confused | 20:33 |
| *** javeriak has quit IRC | 20:33 | |
| *** KLevenstein has quit IRC | 20:37 | |
| *** KLevenstein__ has joined #openstack-ansible | 20:38 | |
| *** KLevenstein__ is now known as KLevenstein | 20:38 | |
| mattt | sigmavirus24: figured i'd do a backport while i go up and down a million times trying to get my daughter to sleep :) | 20:47 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Set tcp_tw_reuse for swift storage hosts https://review.openstack.org/177172 | 20:53 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Add package list and repos to AIO diagnostics https://review.openstack.org/176657 | 20:57 |
| openstackgerrit | Matt Thompson proposed stackforge/os-ansible-deployment: Add package list and repos to AIO diagnostics https://review.openstack.org/177434 | 20:59 |
| *** sacharya has quit IRC | 21:00 | |
| *** britthou_ has quit IRC | 21:01 | |
| *** sdake_ has joined #openstack-ansible | 21:04 | |
| *** Mudpuppy_ has joined #openstack-ansible | 21:05 | |
| *** Mudpuppy_ has quit IRC | 21:07 | |
| *** Mudpuppy_ has joined #openstack-ansible | 21:07 | |
| *** sdake has quit IRC | 21:08 | |
| *** Mudpuppy has quit IRC | 21:08 | |
| *** KLevenstein has quit IRC | 21:14 | |
| *** sdake has joined #openstack-ansible | 21:19 | |
| *** sdake_ has quit IRC | 21:23 | |
| openstackgerrit | Ian Cordasco proposed stackforge/os-ansible-deployment: Lower PyYAML requirement to satisfy Heat https://review.openstack.org/177467 | 21:44 |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Add package list and repos to AIO diagnostics https://review.openstack.org/177434 | 21:47 |
| openstackgerrit | Ian Cordasco proposed stackforge/os-ansible-deployment: Lower requirements pins to satisfy global-requirements https://review.openstack.org/177467 | 21:50 |
| *** stevemar has quit IRC | 21:53 | |
| sigmavirus24 | mancdaz: cloudnull https://review.openstack.org/177467 is the final list of our pins that needed to be changed to satisfy stable/juno g-r | 21:54 |
| *** javeriak has joined #openstack-ansible | 22:05 | |
| *** sdake_ has joined #openstack-ansible | 22:11 | |
| *** sdake has quit IRC | 22:15 | |
| openstackgerrit | Steve Lewis proposed stackforge/os-ansible-deployment: Update Horizon configuration https://review.openstack.org/177473 | 22:17 |
| *** mnestheu1 has quit IRC | 22:19 | |
| *** sdake has joined #openstack-ansible | 22:23 | |
| *** sdake_ has quit IRC | 22:27 | |
| *** sdake_ has joined #openstack-ansible | 22:53 | |
| *** sdake__ has joined #openstack-ansible | 22:55 | |
| *** sdake has quit IRC | 22:56 | |
| *** sdake_ has quit IRC | 22:59 | |
| *** Mudpuppy_ has quit IRC | 23:05 | |
| openstackgerrit | Kevin Carter proposed stackforge/os-ansible-deployment: Run the glance cache cron jobs more frequently https://review.openstack.org/176370 | 23:07 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Set gerrit default branch to kilo https://review.openstack.org/176246 | 23:15 |
| *** sigmavirus24 is now known as sigmavirus24_awa | 23:25 | |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Add package list and repos to AIO diagnostics https://review.openstack.org/177434 | 23:50 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Update Horizon configuration https://review.openstack.org/177473 | 23:50 |
| openstackgerrit | Merged stackforge/os-ansible-deployment: Add basic orchestration tests to the gate check https://review.openstack.org/177021 | 23:50 |
| *** sdake__ has quit IRC | 23:52 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!