| *** karimb has quit IRC | 02:18 | |
| -openstackstatus- NOTICE: Gerrit is going to be restarted due to slow performance | 12:45 | |
| -openstackstatus- NOTICE: Gerrit has been successfully restarted | 12:48 | |
| *** dmsimard|off is now known as dmsimard | 14:02 | |
| *** themurph has joined #ara | 14:03 | |
| dmsimard | boxrick1: saw your messages, sorry I was off yesterday. Did you end up figuring it out ? | 14:18 |
|---|---|---|
| boxrick1 | Yea all good, managed to get it working. | 14:18 |
| boxrick1 | It would be useful to see which host ran a particular playbook since I am using this for about 10 different deployment hosts | 14:19 |
| boxrick1 | With a central ARA box | 14:19 |
| boxrick1 | brb 1 sec gotta nip and grab food from canteen before it shuts! | 14:20 |
| dmsimard | boxrick1: hm, but ARA doesn't know that :) | 14:20 |
| dmsimard | boxrick1: however, I do have an idea about how you could do it -- let me know when you're back :) | 14:20 |
| boxrick1 | Ahh I am back | 14:35 |
| boxrick1 | What are your thoughts then ? | 14:36 |
| dmsimard | boxrick1: so ARA has no clue where it's being run from -- Ansible doesn't really either | 14:39 |
| dmsimard | boxrick1: My idea is that you could have two simple tasks, let me put up a gist, hang on | 14:39 |
| boxrick1 | Is there not a simple case of an extra DB entry ? | 14:39 |
| boxrick1 | Oh by the way, lovely tool. Will come in great use :) | 14:40 |
| dmsimard | boxrick1: you could likely do it this way: https://gist.github.com/dmsimard/a927aebd453ac8f6300f74e6057811cb | 14:41 |
| dmsimard | See docs: http://ara.readthedocs.io/en/latest/usage.html#using-the-ara-record-module | 14:42 |
| dmsimard | You can see what it looks like in practice here at the bottom of the page: http://logs.openstack.org/51/424251/3/gate/gate-ara-integration-latest-ubuntu-xenial/8efa513/logs/build/playbook/149075a7-ba6d-433c-8ce8-821355b6f397/index.html | 14:42 |
| boxrick1 | Ok that makes sense | 14:43 |
| dmsimard | boxrick1: so my design philosophy with ara is to keep it as simple as possible *but* provide the means for people to do what they need, "just an extra DB entry" would make it so there'd be dozens of just another DB entry :) | 14:44 |
| dmsimard | a good example is that ara_record feature came from the need of a user that needed to record which git sha1 the playbook ran from | 14:44 |
| dmsimard | Now, ARA could have gone and checked git history and stuff like that and then record it if there's one available .. but this is much, much more flexible -- and see, it's probably something that would solve your use case as well :) | 14:45 |
| dmsimard | another use case was to make ARA support sending data to collectd/elasticsearch and so on -- ARA doesn't support that but it's CLI provides JSON or YAML output which are good candidates to send fairly raw to other systems if need be | 14:47 |
| boxrick1 | Yea, but I imagine integrating all that would be a fairly monstrous task. So again providing the output to do this makes sense. | 14:48 |
| * dmsimard nods | 14:50 | |
| dmsimard | boxrick1: do let me know if you have any ideas/comments/suggestions, love to get those | 14:50 |
| dmsimard | most of the development is driven by users :) | 14:50 |
| boxrick1 | Well, I managed to blow it up a few times by restarting the database. | 14:51 |
| boxrick1 | So I imagine that if connectivity is patchy then things may become an issue | 14:51 |
| boxrick1 | But so far its great, ill let you know as we use it more. | 14:52 |
| dmsimard | That's actually a good point, I've never tested ara against an unreachable mysql instance .. like, does it bog down/error the playbook execution ? | 14:53 |
| boxrick1 | Oh no, it literally breaks the database. | 14:53 |
| boxrick1 | I had to dump the whole thiong | 14:54 |
| boxrick1 | thing* | 14:54 |
| boxrick1 | I didn't try it mid execution | 14:54 |
| boxrick1 | This was just during fault finding trying to get it up and running, so I probably did more more awefulness which helped it break. | 14:54 |
| dmsimard | I'll create a bug to check that out. | 14:54 |
| boxrick1 | Interestingly that exact scenario has just arisen | 15:08 |
| boxrick1 | I am re-creating ARA container using Ansible, the control host in this case is trying to write to ARA which doesn't exist | 15:09 |
| boxrick1 | And just hangs | 15:09 |
| *** karimb has joined #ara | 15:11 | |
| dmsimard | Hmm, I wonder how we could address that. | 15:12 |
| dmsimard | ara will be "summoned" by ansible due to callback for each task and some other events. I would have to guess sqlalchemy eventually gives up and time outs but then it'll do that a lot of times. | 15:17 |
| dmsimard | Only thing I can think of right now would be to store the entirety of the run in a temporary local dabatase (not unlike what would happen when using SQLite backend) and then only try to push the data once to mysql at the end of the playbook | 15:21 |
| dmsimard | This would lessen the impact on playbook/task execution if there is latency or outage to the remote database server | 15:21 |
| dmsimard | Which sort of makes sense, anyway, I think that if you have a lot of hosts/tasks, just a little latency probably adds up to something significant | 15:22 |
| boxrick1 | I think the best "fix" would be just to fail fast | 16:04 |
| dmsimard | boxrick1: but you'd still fail fast n*y (hosts*tasks) | 16:10 |
| dmsimard | so pretend you fail in 5 seconds instead of, say, 30 seconds | 16:10 |
| dmsimard | I mean, I can make that configurable but it's still pretty painful | 16:10 |
| dmsimard | I'll give it some thought, I think sending all the database transactions in one commit at the end is probably a good bet -- both in terms of reliability (or lack thereof) and in terms of performance/latency | 16:11 |
| dmsimard | mnaser: did that html publisher thing end up working ? | 16:12 |
| mnaser | dmsimard: yess, after playing a bit with the policy settings (i disabled it pretty much) | 16:12 |
| mnaser | we have an ARA report for every build now failing or not failing | 16:13 |
| dmsimard | Neat | 16:13 |
| mnaser | the main reason is time related purposes | 16:13 |
| mnaser | it's helped figure out where things are going slow | 16:13 |
| dmsimard | neat^2 | 16:14 |
| dmsimard | the task result table that's coming in the new version will be sortable so you can sort by duration and even more easily get that :) | 16:14 |
| * dmsimard needs more time | 16:15 | |
| mnaser | yeah i was hoping it would be soratble | 16:15 |
| mnaser | i clicked it and nothing happened, aha :p | 16:15 |
| dmsimard | it's coming | 16:16 |
| dmsimard | turns out it's easier to prototype things than to actually code/implement them | 16:16 |
| dmsimard | who'd have thought | 16:16 |
| dmsimard | jrist: which reminds me, there's no real good spot to put the logo in the new ui ? | 16:17 |
| mnaser | haha | 16:17 |
| mnaser | we've been moving towards horizon | 16:17 |
| mnaser | that's been fun.. | 16:17 |
| jrist | dmsimard: do you have the UI up somewhere that I can look? | 16:17 |
| jrist | or source | 16:17 |
| dmsimard | jrist: nothing to show yet, but say you look at this: http://www.patternfly.org/pattern-library/navigation/horizontal-navigation/horizontal-navigation.html#_ | 16:17 |
| dmsimard | jrist: (I'm getting rid of the left navbar) | 16:17 |
| jrist | yeah so | 16:18 |
| dmsimard | that top black bar there | 16:18 |
| jrist | for tripleo we put it in top | 16:18 |
| jrist | and juts increase height a bit | 16:18 |
| jrist | which is allowed :) | 16:18 |
| dmsimard | the logo needs to fit there in height and it makes the right-most links look meh | 16:18 |
| jrist | indeed | 16:18 |
| dmsimard | because the border doesn't extend all the way down ? | 16:18 |
| dmsimard | jrist: do you have a demo of tripleo-ui somewhere ? | 16:18 |
| jrist | maybe! | 16:18 |
| jrist | not public | 16:18 |
| dmsimard | oh it can be secret between us :P | 16:19 |
| * dmsimard logs on vpn | 16:19 | |
| jrist | https://i.imgur.com/eq41FC4.png | 16:20 |
| dmsimard | hm, when I tried it it made the dividers on the right links not extend all the way down | 16:20 |
| dmsimard | I'll try it again, I guess I had made it too tall | 16:25 |
| *** karimb has quit IRC | 16:26 | |
| *** karimb has joined #ara | 16:29 | |
| boxrick1 | So what does TripleO do then ? | 17:07 |
| *** karimb has quit IRC | 17:55 | |
| *** karimb has joined #ara | 21:40 | |
| *** themurph has quit IRC | 21:58 | |
| *** themurph has joined #ara | 21:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
