| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool feature/zuulv3: Add support for custom ssh port https://review.openstack.org/468752 | 00:57 |
|---|---|---|
| *** bhavik1 has joined #zuul | 04:48 | |
| pabelanger | o/ | 04:48 |
| *** bhavik1 has quit IRC | 06:28 | |
| *** isaacb has joined #zuul | 06:32 | |
| *** hashar has joined #zuul | 08:31 | |
| *** rbergero1 is now known as rbergeron | 08:43 | |
| *** rbergeron has quit IRC | 08:43 | |
| *** rbergeron has joined #zuul | 08:43 | |
| tobiash | fun with ansible and py3 | 09:26 |
| tobiash | when using delegate to localhost it ignores the ansible_python_interpreter setting and uses py3 | 09:27 |
| tobiash | -> synchronize broken | 09:27 |
| tobiash | :( | 09:27 |
| jlk | tobiash: hey we were just talking about this at the Ansible Contributor Summit. | 09:29 |
| tobiash | jlk: yay :) | 09:30 |
| jlk | tobiash: there is some weirdness with which python interpreter gets used on delegate to localhost vs others, and implicit localhost inventory entry vs explicit and whatnot. | 09:30 |
| jlk | it's somewhat known that there are bugs here, because the localhost handling is odd, and has been odd since the beginning. | 09:30 |
| mordred | jlk, tobiash we can grab toshio a little later today... | 09:39 |
| mordred | tobiash: in lib/ansible/inventory/data.py - it says something about setting ansible_python_interpreter explicitly for localhost | 09:43 |
| mordred | tobiash: any chance if we set it explicitly for localhost in addition to setting it more globally it behaves more different? | 09:43 |
| tobiash | mordred: thanks, will have a look on that | 09:44 |
| tobiash | but first have to grab something to eat | 09:44 |
| mordred | tobiash: cool. there's also a note in the syncronize module too... | 09:44 |
| tobiash | mordred: this seems to be fixed in current develop branch: | 09:56 |
| tobiash | https://github.com/ansible/ansible/blob/devel/lib/ansible/inventory/data.py#L86 | 09:56 |
| tobiash | latest release looks broken: | 09:56 |
| tobiash | https://github.com/ansible/ansible/blob/v2.3.1.0-1/lib/ansible/inventory/__init__.py#L507 | 09:56 |
| mordred | tobiash: AHHHH | 09:56 |
| mordred | tobiash: I believeI have latest devel checked out | 09:56 |
| mordred | tobiash: ok - so - toshio says the issue is largely around implicit localhost ... so if we add localhost to our inventory explicitly, it shoudl work properly - even with 2.3.1 | 10:05 |
| tobiash | mordred: trying now | 10:10 |
| tobiash | first try failed as localhost was part of the all group -> plays against all break | 10:11 |
| jlk | jeblair: is there a thing in our test framework that can assert that a traceback didn't happen? I'm trying to add a test for some new code I introduced to limit github reporting for certain events (push events can't manipulate PRs). I can certainly test that the push event did the expected reporting, I want to somehow test that it didn't attempt the unexpected reporting. | 10:18 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Add localhost explicitly to the inventory https://review.openstack.org/476081 | 10:20 |
| mordred | tobiash, jeblair, jlk: ^^ I think that's what we just chatted about with toshio | 10:20 |
| tobiash | mordred: that will add localhost to the all group | 10:21 |
| mordred | tobiash: oh - piddle. yah. that's bad | 10:22 |
| mordred | tobiash: thankyou | 10:22 |
| jlk | oh | 10:22 |
| jlk | uh | 10:22 |
| jlk | crap that's the other shitty thing | 10:22 |
| jlk | implicit localhost is not in all, but explicit is. | 10:22 |
| tobiash | currently trying something similar without putting it into all | 10:22 |
| mordred | you should be able to put it into the group named "ungrouped" | 10:23 |
| jeblair | jlk: i don't think so (especially since that will be deep inside the scheduler -- if it does hit a traceback, the scheduler should just log it and continue). but you can at least check that the reporting didn't happen. | 10:23 |
| jlk | hrm. | 10:24 |
| jlk | I don't think the github driver tracks attempts to report anywhere, so I can't really check the list of things that were reported for lack of a thing. And if there is no PR associated with the push, I can't check that hte PR didn't get reported on, because it never would have been. | 10:25 |
| jeblair | ah, that makes sense | 10:25 |
| mordred | oh - piddle. adding localhost to ungrouped seems to still add it to all :( | 10:25 |
| jeblair | jlk: i don't have a good suggestion right now :( | 10:27 |
| jlk | okay thanks. | 10:30 |
| jlk | another question | 10:30 |
| jlk | in a Fake connection class, can I overload a function (say the report() function) just to do some fake things first, and then still call the parent report() ? | 10:31 |
| jlk | I should be able to do super(FakeGithubConnection, self).report() yeah? | 10:31 |
| mordred | tobiash: ok - ansible core folks say it's not posible to do the explicit inventory fix without localhost going in to all | 10:34 |
| mordred | tobiash: so - it seems our choices are to see if we can extract enough of the fix from devel that it could be safely backported to 2.3 - or figure out a workaround like running ansible-playbook with python2 or something | 10:35 |
| mordred | tobiash: this is fun! :) | 10:35 |
| jeblair | jlk: yeah, i think there are some similar things already. | 10:36 |
| tobiash | mordred: so ungrouped won't work (that was also my idea)? | 10:43 |
| *** ricky_ has joined #zuul | 10:43 | |
| *** ricky_ is now known as Guest85911 | 10:43 | |
| mordred | tobiash: yah - ungrouped still implies all | 10:43 |
| tobiash | crap | 10:43 |
| mordred | tobiash: bcoca is going to look at backporting the specific thing after today | 10:44 |
| mordred | tobiash: he thinks this specific thing might be specific enough to do - devel otherwise has a reworking ofht einventory/variables system | 10:45 |
| *** Guest85911 is now known as ricky_ansible_fe | 10:46 | |
| *** ricky_ansible_fe is now known as rcarrillocruz_fe | 10:46 | |
| *** rcarrillocruz_fe is now known as ricky_fest | 10:46 | |
| *** jkilpatr has quit IRC | 10:49 | |
| *** jkilpatr has joined #zuul | 11:08 | |
| *** ricky_fest has quit IRC | 11:12 | |
| *** isaacb_ has joined #zuul | 11:52 | |
| *** isaacb has quit IRC | 11:53 | |
| *** ricky_fest has joined #zuul | 12:01 | |
| *** isaacb_ has quit IRC | 12:25 | |
| *** isaacb_ has joined #zuul | 12:27 | |
| *** isaacb__ has joined #zuul | 12:40 | |
| *** isaacb__ has quit IRC | 12:40 | |
| *** isaacb_ has quit IRC | 12:42 | |
| tobiash | ok, hacked around it with the find and fetch modules (which seem to work with py3) instead of synchronize for now | 12:46 |
| mordred | tobiash: cool | 13:26 |
| tobiash | mordred: it's much slower and doesn't really work recursively but is ok for now for my use case | 13:27 |
| tobiash | currently crafting a base job which collects system stats with sar on all test nodes during the job | 13:28 |
| tobiash | my problem was copying back the results | 13:28 |
| tobiash | hm, my multi-executor/merger deployment doesn't reliably update all relevant repos prior running the job | 13:43 |
| tobiash | when adding a change in repo a and having a base job in repo b it updates repo b non-deterministically | 13:44 |
| tobiash | (running current tip of v3 branch) | 13:52 |
| *** ricky_fest has quit IRC | 15:10 | |
| SpamapS | I'd support running ansible-playbook with python2 as a workaround until Ansible releases a version that supports py3 for at least the included modules. | 15:41 |
| SpamapS | Zuul is in our control, so py3 is all well and good. But I don't really want to ride the lightning with Ansible. | 15:41 |
| SpamapS | Would mean jumping through some hoops to install ansible rather than having it in requirements.txt | 15:49 |
| *** harlowja has joined #zuul | 17:07 | |
| *** hashar has quit IRC | 17:08 | |
| openstackgerrit | Clark Boylan proposed openstack-infra/nodepool master: Cleanup failed upload records https://review.openstack.org/476223 | 18:04 |
| *** hashar has joined #zuul | 18:05 | |
| openstackgerrit | Merged openstack-infra/nodepool master: Consider hostname in determining local builds https://review.openstack.org/470364 | 18:11 |
| *** hashar has quit IRC | 18:35 | |
| Shrews | that ^^^ reminds me, we should probably think of doing a nodepool feature/zuulv3 branch merge with master soon | 18:39 |
| Shrews | merge FROM master, rather | 18:40 |
| *** hashar has joined #zuul | 18:47 | |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul feature/zuulv3: WIP: Add web-based console log streaming https://review.openstack.org/463353 | 19:02 |
| Shrews | fwiw, trying to test the websocket streaming within a unit test is a herculean task | 19:03 |
| Shrews | damn you, mordred | 19:03 |
| Shrews | :-P | 19:03 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul feature/zuulv3: WIP: Add web-based console log streaming https://review.openstack.org/463353 | 19:05 |
| *** openstackgerrit has quit IRC | 20:03 | |
| *** dmsimard has quit IRC | 20:20 | |
| *** dmsimard has joined #zuul | 20:21 | |
| *** hashar has quit IRC | 20:22 | |
| *** openstackgerrit has joined #zuul | 20:43 | |
| openstackgerrit | Clark Boylan proposed openstack-infra/nodepool master: Cleanup failed upload records https://review.openstack.org/476223 | 20:43 |
| openstackgerrit | Clark Boylan proposed openstack-infra/nodepool master: Disable CleanupWorker thread for test_image_upload_fail https://review.openstack.org/476266 | 20:43 |
| *** jkilpatr has quit IRC | 21:56 | |
| openstackgerrit | Adam Gandelman proposed openstack-infra/zuul feature/zuulv3: Fix parsing of github PR url https://review.openstack.org/476286 | 23:31 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!