dmsimard | dotplus: just saw your question about role duration | 01:49 |
---|---|---|
dmsimard | so I think last I checked the notion of role doesn't get passed down to callbacks, which is what ARA leverages to record data -- the callback doesn't know the relationship between roles and tasks | 01:50 |
dmsimard | closest we have is the file path to the role | 01:50 |
dmsimard | While probably not super convenient, maybe a way to narrow down the duration of a role would be per play | 01:51 |
dmsimard | We have duration for a whole play -- if there is only one role in a play, you'll get the duration for that role effectively | 01:51 |
dmsimard | There is otherwise a "search" function at the task level that is coming eventually | 02:17 |
dmsimard | You can see an early prototype here https://www.youtube.com/watch?v=h3vY87_EWHw | 02:17 |
*** karimb has joined #ara | 08:55 | |
*** permalac has quit IRC | 09:08 | |
*** permalac has joined #ara | 09:19 | |
*** karimb has quit IRC | 09:31 | |
*** permalac has quit IRC | 09:33 | |
*** permalac has joined #ara | 09:35 | |
*** karimb has joined #ara | 09:52 | |
*** karimb has quit IRC | 11:07 | |
*** karimb has joined #ara | 11:35 | |
*** karimb has quit IRC | 12:45 | |
*** karimb has joined #ara | 13:26 | |
*** themurph has joined #ara | 13:58 | |
dotplus | hmm. So the answer is currently, "You can't because Ansible doesn't give role information to callbacks". But I might be able to get that by proxy if I were to search by path (correct if we're careful about role names, which we are) matching 'roles/[org].[rolename]/'. But currently, I would need to that in "post-processing" directly against the DB. | 14:40 |
dotplus | Since the ultimate goal is to be able to understand where the delays/bottlenecks are in our Ansible runs and whether a particular branch/patchset results in a improvement, I'm assuming that we're going to want to make graphs-over-time from this data which probably means getting it into something like ELK. | 14:42 |
dmsimard | dotplus: I asked #ansible-devel this morning about role info being passed down to callbacks but heard nothing but crickets ;) | 14:43 |
dotplus | Perhaps I'm trying to use ARA to hammer a screw and I would be better off writing a separate callback plugin to shovel timing data in to ELK and doing all the processing/graphing in "one" place. Thought? | 14:43 |
dmsimard | dotplus: but yeah, you could probably filter it down either directly through the DB or maybe with the CLI | 14:43 |
dmsimard | FWIW I am very interested in the ELK use case (for my own purposes) but I haven't had the chance to try it out yet. I definitely want to document how to do it. | 14:44 |
dmsimard | dotplus: ara uses the same CLI framework as python-openstackclient (cliff) so you can export CLI results out to JSON, YAML or any other format that might be more convenient to re-use with systems like ELK | 14:45 |
dotplus | unfortunately this is apparently a low priority if not quite exactly skunkworks, so I can't give much time (to my chagrin). But if I get any useful progress on Ansible timing data -> ELK, I'll let you know. | 14:46 |
dmsimard | dotplus: np, it's something I want/need to work on in the short term so I'll let you know as well if I figure something out | 14:47 |
dotplus | yeah, I recognised the output stuff in the --help, I'm familiar with python-openstackclient. | 14:47 |
dmsimard | dotplus: in the meantime, I don't know how possible it is for you but like I said, a workaround would be to have a one-to-one role/play relationship | 14:47 |
dmsimard | so that all tasks are grouped under one play and easily filterable out of the box | 14:48 |
dotplus | no, I have no control over that. It's already grown *far* beyond that sort of thing. | 14:48 |
dmsimard | okay, because for instance in http://logs.openstack.org/11/429011/4/check/gate-ara-integration-latest-centos-7/b8777f2/logs/build/playbook/9a6e7b6c-24c4-40e2-b254-9dc32af7d6a0/index.html there's two plays with their respective durations | 14:48 |
dotplus | yeah, I wish. | 14:51 |
dotplus | btw, links to http://logs.openstack.org/11/429011/4/check/gate-ara-integration-* render completely unstyled for me. Is that expected? | 14:52 |
dmsimard | Hm, works for me. What browser/OS are you reproducing it on ? | 14:53 |
dotplus | firefox/OSX. I'll check something else | 14:55 |
dotplus | ok, weird. works in chrome. | 14:55 |
dotplus | meh, this is why I hate graphical anything. | 14:56 |
dotplus | never mind, I'll just use chrome for this. don't waste an iota of your precious brain cycles on this. | 14:56 |
dmsimard | dotplus: there's some .gz encoding magic going on, since logs.o.o compresses things automatically | 15:14 |
dmsimard | perhaps it's related to that. | 15:14 |
boxrick1 | dmsimard: I have a question regarding ansible.cfg Priorities and ARA | 15:43 |
dmsimard | boxrick1: shoot | 15:43 |
boxrick1 | I have been deploying it using ENV vars, which works fine for most of my stuff | 15:43 |
boxrick1 | However when using Ansible-Openstack it requires a number of its own plugins. | 15:44 |
boxrick1 | And this all breaks if I have ARA configured using env vars | 15:44 |
boxrick1 | Sadly they cannot live alongside each other since the ansible-openstack stuff blatts over the .cfg file with its own stuff | 15:44 |
boxrick1 | Any suggestions ? | 15:44 |
boxrick1 | I almost need the configs to merge. | 15:45 |
dmsimard | boxrick1: openstack-ansible knows about ARA so they might have some tips -- I've never personally used it, but it looks like they setup a default callback path in /etc/ansible, i.e: https://review.openstack.org/#/c/396324/5/scripts/gate-check-commit.sh | 15:45 |
dmsimard | boxrick1: it looks like in https://review.openstack.org/#/c/396565/11/test-ansible-env-prep.sh they symlink the ara callback plugin dir to the default dir | 15:46 |
boxrick1 | So its almost as if it is already catered for. | 15:47 |
dmsimard | well my understanding is that those are gate-specific | 15:48 |
dmsimard | but maybe you could do something similar | 15:48 |
dmsimard | in an actual deploymenty | 15:48 |
boxrick1 | Both of those are different options that help me get around the problem though :) | 15:53 |
boxrick1 | Cheers | 15:53 |
dmsimard | woohoo :) | 15:54 |
dmsimard | any other questions ? :P | 15:54 |
boxrick1 | Sure while you are here, hows your new interface going? | 15:54 |
dmsimard | pretty bad. | 15:55 |
dmsimard | I need help from jrist for just a silly first step of it but in reality we're both busy with the upcoming release of Ocata | 15:55 |
jrist | confirmed | 15:56 |
jrist | I need to load up all of the ara bits in a test env | 15:56 |
jrist | so I can load the UI :) | 15:56 |
dmsimard | and to make matters worse I'm going down to Boston this week T_T | 15:56 |
dmsimard | well, Westford | 15:56 |
dmsimard | jrist: you happen to be in westford office by chance? :p | 15:56 |
boxrick1 | So what do you guys both do full time then? | 15:57 |
dmsimard | I'm part of the RDO engineering team at Red Hat, I help build, package, test and ship things. | 15:58 |
dmsimard | ARA is a side/personal-ish project that I want to help me in my job | 15:58 |
dmsimard | A lot of our tooling and CI is based on Ansible so making things easier to troubleshoot is one of my goals | 16:00 |
boxrick1 | Ah cool, fair enough. That all makes sense. Big fan of Redhat, sadly where I am right now favours Ubuntu. | 16:03 |
jrist | dmsimard: nope, remote | 16:06 |
jrist | boxrick1: one of the leads on tripleo-ui team | 16:07 |
dotplus | boxrick1: If I understand what you're saying, you need to tell Ansible to look in multiple paths for plugins? you can do things like "callback_plugins = /usr/share/ansible/plugins/callback:/path/to/ansible-openstack/plugins/callbacks:$VIRTUAL_ENV/lib/python2.7/site-packages/ara/plugins/callbacks" | 16:39 |
boxrick1 | Yea thats what I ended up doing | 16:39 |
boxrick1 | I have a role which deploys ARA, then later on something else tried to add its own plugins path. | 16:39 |
boxrick1 | So I just define them both now within my role using ^^^ so it isn't upset. | 16:40 |
dotplus | yup. we have similar constraints. | 16:40 |
*** cliles has quit IRC | 16:55 | |
*** cliles has joined #ara | 16:56 | |
openstackgerrit | David Moreau Simard proposed openstack/ara master: Add new ara_record types: list and dict https://review.openstack.org/429011 | 17:24 |
*** karimb has quit IRC | 17:36 | |
*** Ravi_ has joined #ara | 18:39 | |
*** Ravi_ has quit IRC | 19:56 | |
*** karimb has joined #ara | 22:40 | |
*** karimb has quit IRC | 23:40 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!