| -openstackstatus- NOTICE: The logs.openstack.org filesystem has been restored to full health. We are attempting to keep logs uploaded between the prior alert and this one, however if your job logs are missing please issue a recheck. | 04:46 | |
| *** ChanServ changes topic to "The logs.openstack.org filesystem has been restored to full health. We are attempting to keep logs uploaded between the prior alert and this one, however if your job logs are missing please issue a recheck." | 04:46 | |
| *** ChanServ changes topic to "OpenStack & Chef | chef.io/openstack | bit.ly/os-chef-wiki | bit.ly/os-gerrit-inbox | bit.ly/os-chef-repo | bit.ly/os-chef-meetings-logs | bit.ly/os-chef-meeting-agendas | bit.ly/stackalytics-chef" | 04:52 | |
| *** os-chef-bot has quit IRC | 06:25 | |
| *** os-chef-bot has joined #openstack-chef | 07:40 | |
| openstackgerrit | Merged openstack/cookbook-openstack-ops-database master: ops-database refactor for Pike and Chef 13 https://review.openstack.org/523050 | 13:13 |
|---|---|---|
| openstackgerrit | Merged openstack/cookbook-openstack-ops-database master: Add receipes for mariadb cluster deployment https://review.openstack.org/529819 | 13:13 |
| *** openstackgerrit has quit IRC | 13:18 | |
| sc` | my dev lab won't be back in full function until early next month, when i hope to have more time to put forth to queens | 14:07 |
| sc` | frickler: from my own testing on bare metal, package_overrides existing did nothing one way or the other so long as the apt.conf.d file was emitted. initially, i blanked them all out | 14:37 |
| sc` | i had more trouble out of the attribute until i made it a config, between chef 12 and 13 | 14:38 |
| sc` | i meant to direct that to jklare | 14:39 |
| sc` | E_NOSLEEP | 14:39 |
| sc` | jklare: i'm inclined to just delete the whole conditional if nothing is using it. apt is quite happy with the config file in testing | 14:41 |
| sc` | all of this madness with apt.conf.d and package_overrides came about from trying to drop old cookbook cruft, like apt since it was rolled into chef-client | 14:45 |
| sc` | both apt and yum are in chef-client as of late 12, so including them is more misleading than excising | 14:46 |
| sc` | people would expect include_recipe 'apt' to work because it's there, but it does nothing now. it used to run apt-get -q update | 14:47 |
| jklare | sc` i am a little confused now | 15:20 |
| sc` | i know. i was, too. see my comments on 523030 | 15:21 |
| sc` | both apt cache updating and package installation had problems between chef 12 and 13. one of the sticking points was single versus double quotes in ruby, and switching chef versions made that difficult to debug and converge | 15:22 |
| sc` | the patches as they sit in gerrit attempt to make apt easier to maintain and debug | 15:23 |
| sc` | all of my 'refactor' changes contain about two or three years' worth of chef deprecations, because of how 12 to 13 was | 15:26 |
| sc` | i want to get kitchen working as a third-party ci | 15:27 |
| sc` | i feel like that's our path to delightful | 15:27 |
| sc` | the outcome of all of these changes is that the cookbooks will test with tempest for every chef run, be it openstack ci or local testing | 15:42 |
| jklare | sc` still confused | 15:47 |
| sc` | about? | 15:47 |
| jklare | sc` wouldnt it be easier to use these attributes? https://github.com/chef-cookbooks/apt/blob/v6.1.4/templates/10dpkg-options.erb | 15:47 |
| jklare | and bump the apt version to 6 | 15:47 |
| sc` | if that cookbook actually did something, sure. however, apt and yum are rolled up into chef core | 15:47 |
| jklare | maybe i missed the announcement about the apt cookbook beeing deprecated | 15:48 |
| jklare | i mean a big part of the apt libs were moved to chef core | 15:49 |
| jklare | but not everything | 15:49 |
| jklare | e.g. these config files | 15:49 |
| jklare | or the whole configuration and unattended upgrades part | 15:49 |
| jklare | as well as the apt_preferences resource | 15:50 |
| jklare | is that all going to be included in chef13? | 15:50 |
| sc` | it is in chef 13, yes | 15:50 |
| sc` | https://docs.chef.io/resource_apt_preference.html | 15:50 |
| sc` | apt_preferences was included in 13.3 | 15:50 |
| sc` | 2.3.4 is what i developed on, and it has at least 13.4 | 15:51 |
| sc` | chefdk 2.3.4 that is | 15:51 |
| jklare | ah ok | 15:52 |
| sc` | it was confusing to me until i had all of the changes staged locally | 15:52 |
| sc` | i'm going to admit: i have a lot of wip cruft where i would attack the problem from different angles, until i came to what i proposed to gerrit | 15:54 |
| sc` | there are a few weeks' worth of now throwaway changes here and there | 15:55 |
| sc` | but that's how i derived on what i proposed | 15:55 |
| sc` | node.foo.bar => node['foo']['bar'] was the most widespread, and obscured a lot of visibility with tooling | 15:57 |
| sc` | https://github.com/chef/chef/tree/master/lib/chef/resource has apt_package, apt_preference, apt_repository, apt_update, yum_repository and yum_package | 15:59 |
| jklare | ok | 16:00 |
| jklare | but as long as any of our cookbooks (or the dependencies) include the apt cookbook, we will generate 2 files containing apt dpkg options in /etc/apt/apt.conf.d/ | 16:01 |
| jklare | which might also cause some trouble | 16:01 |
| sc` | then that would be suboptimal | 16:01 |
| jklare | not sure how these files are merged | 16:01 |
| jklare | so i would think we try to remove the "include_recipe 'apt'" from out cookbooks in favor of the apt_update resource | 16:02 |
| jklare | but for the dpkg config i would prefere to use the attributes from the apt cookbook | 16:03 |
| sc` | i've looked through all 12 cookbooks and the only one that does include it is openstack-common::default | 16:03 |
| sc` | 523030 replaces that with apt_update | 16:04 |
| jklare | ok | 16:04 |
| sc` | the way that chef 13 includes the resources, the corresponding ones in the apt and yum cookbooks no-op | 16:05 |
| sc` | ~ | 16:07 |
| sc` | hi kitty | 16:07 |
| jklare | so after removing apt from the metadata there, does your Berksfile.lock in the test env still include the apt cookbook? | 16:17 |
| jklare | so does someone else pull it in? | 16:17 |
| sc` | mariadb pulls it in, but mariadb is overdue for some love | 16:17 |
| sc` | i checked the dep graph after all changes were staged | 16:17 |
| sc` | the end outcome is to remove build-essential, in the chef community | 16:18 |
| sc` | https://www.dropbox.com/s/pr6hdqnnm7belih/chef_openstack_20171112_noconstraints.png?dl=0 per the last berks viz i ran (this is roughly when i was able to converge without depending on apt or yum directly) | 16:24 |
| sc` | ceph and mariadb are the ones that pull in apt | 16:25 |
| sc` | that's actually a bit stale, since i removed ceph from openstack-block-storage. i meant to open a blueprint for changing it to the more alive version | 16:27 |
| sc` | https://github.com/ceph/ceph-chef is the alive fork of the ceph cookbook | 16:27 |
| sc` | of course, even there, it's not chef 13 ready | 16:27 |
| sc` | one indicator is metadata.rb - current fc will want a slightly different syntax for the license string | 16:29 |
| sc` | it won't even converge on a chef 13 client due to method versus bracket syntax, so no sense in including it, or any of the cookbooks that stopped working with chef 13 | 16:33 |
| jklare | :) | 16:33 |
| jklare | we use our own fork and have recently patched it for chef 13 | 16:34 |
| jklare | https://github.com/cloudbau/ceph-cookbook | 16:34 |
| jklare | but there are a lot of patches in this one, since the original one has not been maintained for quite some time | 16:34 |
| jklare | i will check if we can push the chef13 changes to the one you mentioned | 16:35 |
| sc` | https://www.dropbox.com/s/5trfhlis1yumtmg/chef_openstack_20180115.png?dl=0 here's how the dep graph looks as of this morning, including the changes in gerrit | 16:39 |
| sc` | compute still has a depends on the old ceph cookbook | 16:40 |
| sc` | mariadb is the one that needs the most work to clean up the graph | 16:44 |
| jklare | yeah, i think the mariadb cookbook is just slightly better maintained than the percona one :) | 16:47 |
| sc` | it has more recent eyes, at least | 16:47 |
| sc` | i'm basing that on activity less than a week ago on https://github.com/sous-chefs/mariadb/issues/186 | 16:48 |
| sc` | it's an acknowledgement, but activity regardless | 16:48 |
| sc` | i'm holding myself from that rabbit hole because it will certainly overextend me to cover that cookbook in addition to the other things i do | 16:49 |
| sc` | maintaining cookbooks is basically my side piece. it doesn't get the utmost priority | 16:53 |
| *** openstackgerrit has joined #openstack-chef | 17:05 | |
| openstackgerrit | Samuel Cassiba proposed openstack/cookbook-openstack-compute master: Dropped legacy Ceph cookbook https://review.openstack.org/533740 | 17:05 |
| sc` | another thing i'm doing is trying to make it so the cookbooks come from supermarket for the point of entry, and not from a bunch of local git repos with sparse breadcrumbs | 17:10 |
| sc` | kitchen + supermarket makes kicking the tires easy, which is what's needed in openstack at large | 17:12 |
| sc` | extending that to third-party ci of some sort, it makes automated scenario-based testing less painful | 17:20 |
| sc` | by that same token, pipelining can eventually be set up using stove to deliver updates directly to supermarket. i don't imagine it's a huge amount of code to do it, but haven't allocated time to it because of chef 13 | 17:25 |
| sc` | the added bonus is that it becomes a measure of activity that one can follow | 17:26 |
| -openstackstatus- NOTICE: Zuul has been restarted and has lost queue contents; changes in progress will need to be rechecked. | 18:22 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!
