| *** jamielennox|away is now known as jamielennox | 00:08 | |
| SpamapS | jeblair: http://paste.openstack.org/show/591877/ <-- does this look valid to you? I'm not sure I truly understand how auth: is supposed to be used. :-P | 00:16 |
|---|---|---|
| SpamapS | jeblair: when we get into the launcher, the auth sections are not attached to the jobs that run. | 00:16 |
| SpamapS | ergo, no swift params are sent | 00:16 |
| *** harlowja_ has quit IRC | 00:32 | |
| SpamapS | jeblair: more for the plot.. if I add 'inherit: true' to the _project_ job definitions, it works. (Which seems backwards to me) | 00:36 |
| * SpamapS runs into "oh and we have to configure the connection" and EOD's | 00:40 | |
| *** harlowja has joined #zuul | 00:51 | |
| *** jamielennox is now known as jamielennox|away | 01:04 | |
| openstackgerrit | dongwenjuan proposed openstack-infra/nodepool: add log input para to ssh_connect https://review.openstack.org/408873 | 01:11 |
| *** jamielennox|away is now known as jamielennox | 01:20 | |
| *** saneax is now known as saneax-_-|AFK | 01:30 | |
| *** harlowja has quit IRC | 02:55 | |
| ianw | jeblair greghaynes pabelanger: got in contact with fedora-infra at #fedora-admin and they fixed up an issue with the mirrors and stale data. so think this one was out of our hands | 03:28 |
| greghaynes | ianw: ah, thanks | 03:29 |
| openstackgerrit | Joshua Hesketh proposed openstack-infra/zuul: Add note about redundant file https://review.openstack.org/408949 | 04:42 |
| openstackgerrit | Joshua Hesketh proposed openstack-infra/zuul: Don't try to execute _stop commands https://review.openstack.org/408950 | 04:42 |
| openstackgerrit | Joshua Hesketh proposed openstack-infra/zuul: Add command list into server.py https://review.openstack.org/408951 | 04:42 |
| openstackgerrit | Joshua Hesketh proposed openstack-infra/zuul: Add zmq back into launcher server https://review.openstack.org/408952 | 04:42 |
| *** saneax-_-|AFK is now known as saneax | 05:46 | |
| *** hashar has joined #zuul | 08:21 | |
| *** yolanda has quit IRC | 09:22 | |
| *** Cibo_ has quit IRC | 10:44 | |
| *** Cibo_ has joined #zuul | 10:46 | |
| *** bhavik1 has joined #zuul | 11:31 | |
| *** hashar is now known as hasharLunch | 12:18 | |
| *** jamielennox is now known as jamielennox|away | 12:33 | |
| *** hasharLunch is now known as hashar | 13:07 | |
| *** abregman has joined #zuul | 13:17 | |
| *** abregman has quit IRC | 13:17 | |
| *** patrickeast has quit IRC | 14:38 | |
| *** patrickeast has joined #zuul | 14:38 | |
| *** saneax is now known as saneax-_-|AFK | 14:52 | |
| *** zaro has quit IRC | 15:02 | |
| *** zaro has joined #zuul | 15:03 | |
| *** docaedo has left #zuul | 15:22 | |
| jeblair | SpamapS: http://paste.openstack.org/show/591877/ looks right enough. there isn't much auth-related code in v3 yet, so yeah, there will be configuration parsing/model stuff to write (including the inherit thing) as well as hooking that into ansible variables in the launcher. you may want to check in with rcarrillocruz who is working on the 'secrets' system which is also part of auth, though i don't think he has touched that part of it yet (is ... | 16:56 |
| jeblair | ... mostly working on encryption keys, etc). | 16:56 |
| SpamapS | jeblair: I assume the secrets part will also include something to setup swift connections. | 17:03 |
| jeblair | SpamapS: no, that's separate -- 'secrets' here referring just to user supplied key/value secrets. | 17:08 |
| *** hashar has quit IRC | 17:31 | |
| *** bhavik1 has quit IRC | 17:32 | |
| SpamapS | jeblair: Ah, k. Swift is more-specialer then :- | 18:00 |
| SpamapS | :-P | 18:00 |
| jeblair | SpamapS: yeah, we probably want to build a formal plugin interface for auth plugins, but we don't have one yet. maybe we can do that for 3.0; or maybe for 3.1. :) | 18:01 |
| SpamapS | Yeah I think when the third one shows up is the time to design that interface. :) | 18:06 |
| SpamapS | I'm not going to be around consistently today.. taking care of a sick baby (who is about to wake up from morning nap)_ | 18:07 |
| Shrews | jeblair: 408808 actually has a redundant check that i just noticed. it's safe, though | 18:17 |
| jeblair | Shrews: oh yeah... we should actually probably read the object again inside the lock | 18:17 |
| Shrews | jeblair: well, the idea was to avoid unnecessary zk ops, so that would add an extra op. | 18:18 |
| Shrews | i think it's better to just remove it | 18:18 |
| openstackgerrit | Merged openstack-infra/nodepool: Add __repr__ to ConfigValue objects https://review.openstack.org/408776 | 18:19 |
| Shrews | only on rare occasions would it set it to DELETING unnecessarily | 18:19 |
| openstackgerrit | Merged openstack-infra/nodepool: Check for in progress build/upload in CLI https://review.openstack.org/408794 | 18:19 |
| jeblair | Shrews: yeah, i think removing the check is fine. | 18:20 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool: Re-use build data when we set for DELETING https://review.openstack.org/408808 | 18:21 |
| Shrews | jeblair: there ya go | 18:21 |
| Shrews | mordred: could you poke 408808 once more? | 18:24 |
| mordred | Shrews: done | 18:26 |
| *** harlowja has joined #zuul | 18:32 | |
| *** harlowja_ has joined #zuul | 18:34 | |
| *** harlowja has quit IRC | 18:35 | |
| openstackgerrit | Merged openstack-infra/nodepool: Re-use build data when we set for DELETING https://review.openstack.org/408808 | 19:28 |
| *** openstack has joined #zuul | 19:47 | |
| pabelanger | both builders have been restarted | 19:52 |
| pabelanger | had to do the touch fedora-24-0000000005.raw trick again to remove pending deleting diskimages | 19:52 |
| pabelanger | going to write up a test for that now | 19:52 |
| pabelanger | make sure it works properly | 19:53 |
| clarkb | pabelanger: need to erap the deletes in if os.path.exists ? | 19:55 |
| pabelanger | clarkb: not sure, but this happens when we stop nodepool-builder while a diskimage build is in progress. Should be easy to reproduce and see why it happens | 19:56 |
| openstackgerrit | Paul Belanger proposed openstack-infra/nodepool: Make diskimage-builder command configurable for testing https://review.openstack.org/404976 | 20:43 |
| openstackgerrit | Paul Belanger proposed openstack-infra/nodepool: Properly cleanup failed diskimage builds https://review.openstack.org/409327 | 20:43 |
| pabelanger | clarkb: Shrews: ^ reproduced our issue, test included | 20:43 |
| mordred | pabelanger: woot! | 20:44 |
| Shrews | pabelanger: that won't work | 20:46 |
| Shrews | pabelanger: we don't want to delete the zk node unless the local file is deleted first | 20:46 |
| Shrews | pabelanger: otherwise, a builder that doesn't own the build could delete the zk node, and the owner would never know to delete the local files | 20:47 |
| clarkb | maybe we need to store the builder id in with the build data, and if self == builder_id and file not on disk then remove? | 20:50 |
| clarkb | that still leaves you open to the issue where a builder goes away completely but that will be less common | 20:50 |
| Shrews | i'm not sure that i'm clear on the issue... we stop the builder, and the build is recored as FAILED, right? | 20:51 |
| Shrews | oh, now i remember. it gets to the DELETING state and never goes away | 20:52 |
| Shrews | clarkb: we used to do that check, but jeblair suggested we remove that for some reason i do not recall now | 20:53 |
| clarkb | Shrews: correct its set to delete but it can't delete because there is no file on disk | 20:54 |
| clarkb | another way to address it would be to touch empty files at the start of every build and let dib overwrite them | 20:54 |
| clarkb | but that way you have the markers on disk for when a failure happens | 20:54 |
| clarkb | thats a fairly simple fix if we want to make it "symmetric" | 20:55 |
| jeblair | clarkb, Shrews: i think the builder name check is what we'll have to do (i think we discussed this already?) | 20:56 |
| *** jamielennox|away is now known as jamielennox | 20:58 | |
| pabelanger | Shrews: yes, DELETING state, doesn't go away until we touch a fake file | 20:58 |
| pabelanger | clarkb: not sure I'm a fan of the touch file and have dib overwrite | 21:02 |
| jeblair | can someone go ahead and implement the builder name check? | 21:05 |
| jeblair | here's where we talked about this yesterday: http://eavesdrop.openstack.org/irclogs/%23zuul/%23zuul.2016-12-08.log.html#t2016-12-08T20:10:30 | 21:05 |
| Shrews | pabelanger: in deleteLocalBuild, if the "if not files" is true, compare hostnames there. you may need to pass in the builder in | 21:07 |
| Shrews | just return True if the names match | 21:08 |
| Shrews | s/pass in the builder in/pass in the builder name/ | 21:08 |
| Shrews | or the build object | 21:08 |
| pabelanger | sure, let me poke at it | 21:09 |
| Shrews | pabelanger: self._hostname should already be set for you | 21:11 |
| openstackgerrit | Paul Belanger proposed openstack-infra/nodepool: Properly cleanup failed diskimage builds https://review.openstack.org/409327 | 21:25 |
| pabelanger | Shrews: jeblair: If I understand the request correctly^ | 21:26 |
| jeblair | pabelanger: can you avoid moving the log message? it's going to false fire on non-responsible builders | 21:29 |
| Shrews | pabelanger: +1, but good point jeblair just made | 21:29 |
| pabelanger | yes, let me move back | 21:30 |
| openstackgerrit | Paul Belanger proposed openstack-infra/nodepool: Properly cleanup failed diskimage builds https://review.openstack.org/409327 | 21:32 |
| pabelanger | jeblair: when you have time too: 404976 moves fake-image-create out of builder.py too | 21:52 |
| *** jamielennox is now known as jamielennox|away | 22:04 | |
| *** jamielennox|away is now known as jamielennox | 22:11 | |
| *** jesusaur has quit IRC | 22:15 | |
| adam_g | test suite question: a FakeChange that has its data['status'] = 'MERGED' should be merged into the corresponding repo in /tmp/$tmpdir/zuul-test/upstream/ ? | 22:42 |
| adam_g | or do they not actually hit the merger? | 22:43 |
| adam_g | (v3, btw) | 22:43 |
| *** cinerama has quit IRC | 22:47 | |
| *** cinerama has joined #zuul | 22:47 | |
| jeblair | adam_g: i *think* so, but that's a relatively recent change | 23:31 |
| adam_g | jeblair: i guess what i should be asking is whether post-gate merging expected to be functional at all ATM? | 23:36 |
| adam_g | ive found what i think is a bug in the launcher merging | 23:36 |
| adam_g | and was trying to trace a change end-to-end, and diff'ing behavior against a v2.5 env | 23:37 |
| jeblair | adam_g: yeah, FakeChange.setMerged() updates the upstream git repo, and it's called by FakeGerritConnection.review() (which is the thing that zuul does to tell gerrit to merge something) | 23:38 |
| jeblair | adam_g: so it *should* dtrt to the upstream repo. | 23:39 |
| jeblair | adam_g: oh but wait, you said 'launcher merging' | 23:39 |
| jeblair | adam_g: which makes me think you're actually looking at a different subsystem | 23:39 |
| adam_g | jeblair: well, the launcher merging thing was something else i was looking at | 23:40 |
| jeblair | adam_g: the merging that happens in the launcher is the speculative merging -- it should only modify the local repo in the JobDir that's created for each job | 23:40 |
| adam_g | but got me trying to setup a basic end-to-end test, trying to assert things get merged upstream in the correct order | 23:40 |
| adam_g | jeblair: right | 23:40 |
| jeblair | adam_g: ok cool. btw, there are some tests in test_scheduler that are similar to what you describe, they may be helpful | 23:41 |
| adam_g | let me push what ive got, cause ive got another question while i have you | 23:41 |
| jeblair | adam_g: test_parallel_changes is a basic A,B,C test -- though it only checks the change metadata, it doesn't actually inspect the repo (but there should be others that do) | 23:42 |
| openstackgerrit | Adam Gandelman proposed openstack-infra/zuul: Re-enable test_build_configuration_conflict https://review.openstack.org/409376 | 23:48 |
| adam_g | jeblair: so looking through self.history (on the test in ^), i see the correct number of jobs being run, but not with the expected dependent changes included | 23:50 |
| adam_g | http://paste.openstack.org/show/591994/ | 23:51 |
| adam_g | i'd expect one of those merge tests to include 3,1 + 1,2 | 23:52 |
| adam_g | hmm, ill look closer at test_parallel_changes, that appears to be producing the result i'd expect | 23:54 |
| jeblair | adam_g: i think there's going to be a v2->v3 difference here because of the merge conflict | 23:56 |
| adam_g | ohh | 23:56 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!