| *** kmalloc has quit IRC | 00:29 | |
| tristanC | corvus: would you mind replacing the V3 development section with the relevant storyboard link in the chan topic? | 01:05 |
|---|---|---|
| tristanC | corvus: and is ansible-2.4 support in the scope of the 3.0 release? | 01:06 |
| SpamapS | I thought we had already removed the cap | 01:09 |
| tristanC | The patches are still in flight, i added this topic: https://review.openstack.org/#/q/topic:zuul-ansible-2.4 | 01:14 |
| mordred | tristanC, SpamapS: I think we're waiting on figuring out how to roll out the transition for openstack | 01:31 |
| *** rlandy has quit IRC | 01:32 | |
| *** haint_ has joined #zuul | 01:51 | |
| *** haint93 has quit IRC | 01:55 | |
| *** harlowja has quit IRC | 02:15 | |
| *** JasonCL has quit IRC | 03:19 | |
| *** mgagne has quit IRC | 03:22 | |
| *** jamielennox has quit IRC | 03:23 | |
| *** mgagne has joined #zuul | 03:24 | |
| *** mgagne is now known as Guest87240 | 03:24 | |
| *** jamielennox has joined #zuul | 03:29 | |
| *** JasonCL has joined #zuul | 03:38 | |
| *** JasonCL has quit IRC | 03:39 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool master: Implement a static driver for Nodepool https://review.openstack.org/535553 | 03:59 |
| *** bhavik1 has joined #zuul | 04:03 | |
| *** bhavik1 has quit IRC | 04:14 | |
| Wei_Liu | tristanC: I implement a task to send the signal to scheduler pid to make it reload the tenant config with command "kill -HUP `cat /var/lib/zuul/run/zuul-scheduler.pid`", but it failed due to "No such process". Bubblewrap is used in my executor, so does it mean I can not kill a process out of bubblewrap? | 04:18 |
| tristanC | Wei_Liu: indeed, bubblewrap doesn't give access to the host processes. | 04:21 |
| tristanC | Wei_Liu: since you may want to scale executor on dedicated instance, perhaps you can setup ssh access and use "ssh scheduler kill -HUP"... | 04:22 |
| SpamapS | Wei_Liu: I suggest ssh'ing back into the scheduler. They aren't intended to be on the same box usually. | 04:22 |
| *** elyezer has quit IRC | 04:34 | |
| *** elyezer has joined #zuul | 04:37 | |
| *** elyezer has quit IRC | 04:44 | |
| *** elyezer has joined #zuul | 04:49 | |
| *** harlowja has joined #zuul | 04:51 | |
| *** elyezer has quit IRC | 04:55 | |
| *** elyezer has joined #zuul | 04:57 | |
| Wei_Liu | tristanC, SpampaS, Thanks | 05:13 |
| *** threestrands has quit IRC | 05:51 | |
| *** xinliang has quit IRC | 05:57 | |
| *** threestrands has joined #zuul | 06:01 | |
| *** threestrands has quit IRC | 06:01 | |
| *** threestrands has joined #zuul | 06:01 | |
| *** xinliang has joined #zuul | 06:09 | |
| *** xinliang has quit IRC | 06:09 | |
| *** xinliang has joined #zuul | 06:09 | |
| *** xinliang has quit IRC | 06:21 | |
| *** xinliang has joined #zuul | 06:21 | |
| Wei_Liu | tristanC, SpamapS: wow, it worked, Thanks a lot. One more thing, can I use the same way to update the site-variables.yaml? | 06:30 |
| tristanC | Wei_Liu: what do you mean by site-variables.yaml? | 06:31 |
| SpamapS | That's a config item | 06:31 |
| SpamapS | variables that can be set on every job | 06:32 |
| SpamapS | useful for overriding defaults in zuul-jobs | 06:32 |
| tristanC | hum, if it's in git repository, then maybe you could git push? | 06:32 |
| Wei_Liu | tristanC, yes | 06:33 |
| Wei_Liu | SpamapS: sometimes, I want to publish some global variables to my zuul jobs, I will use the site-variables.yaml, | 06:35 |
| SpamapS | Yes that's what it is for. | 06:37 |
| SpamapS | Wei_Liu: although, if you just want global variables for a single tenant, you can set them on the base job. | 06:37 |
| SpamapS | which might be simpler and easier for users to discover | 06:38 |
| tobiash | SpamapS: did you already repropose the memory governor to master? | 06:46 |
| tobiash | I cannot find it | 06:46 |
| tobiash | SpamapS: do you mind if I cherry pick this to master? | 06:50 |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Add memory awareness to system load governor https://review.openstack.org/539426 | 06:54 |
| tobiash | SpamapS: cherry picked and restored my last vote on that ^^ | 06:54 |
| *** elyezer has quit IRC | 06:57 | |
| *** harlowja has quit IRC | 07:01 | |
| *** elyezer has joined #zuul | 07:01 | |
| SpamapS | tobiash: I did not, pabelanger was taking it over | 07:02 |
| SpamapS | tobiash: glad you're picking it up. :) | 07:02 |
| tobiash | SpamapS: ok :) | 07:02 |
| * SpamapS just hasn't had the time :-P | 07:02 | |
| tobiash | no problem, just wondered where that landed | 07:02 |
| *** elyezer has quit IRC | 07:08 | |
| *** elyezer has joined #zuul | 07:12 | |
| *** jpena|off is now known as jpena | 08:43 | |
| *** elyezer has quit IRC | 08:57 | |
| *** elyezer has joined #zuul | 08:58 | |
| *** elyezer has quit IRC | 10:02 | |
| *** elyezer has joined #zuul | 10:14 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Fix statsd documentation about events https://review.openstack.org/539482 | 10:18 |
| *** sshnaidm|afk is now known as sshnaidm | 10:56 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/nodepool master: [WIP] Allow operator to set held node's TTL https://review.openstack.org/539493 | 11:00 |
| *** threestrands has quit IRC | 12:08 | |
| andreaf | tobiash, mordred: will zuul only merge dict variables from jobs? If I have a list variable and I what to enable children jobs to extend the list, do I have to make the list a dict with some fake values? | 12:40 |
| andreaf | tobiash, mordred: the specific case is the list of extension of files to be renamed to _txt in stage-output - the list is made of unique elements so a dict would fix but having fake values would be a bit odd | 12:42 |
| tobiash | andreaf: I think that's in line with default behavior in ansible which by default overwrites top level variables | 12:43 |
| tobiash | otherwise one would need to implement tree merging which could have unwanted side effects for others expecting vars to overwrite each other | 12:43 |
| tobiash | but I agree that your use case seems not to be fully supported yet | 12:44 |
| tobiash | andreaf: but you also could do this slightly different | 12:45 |
| andreaf | tobiash: thanks - how | 12:45 |
| andreaf | ? | 12:45 |
| tobiash | like having a myfoofiles_<jobname> variable which is a list | 12:46 |
| tobiash | and add several of these in your inheritance structure | 12:46 |
| tobiash | you could merge them in ansible then | 12:46 |
| tobiash | but I guess that's more complicated | 12:47 |
| andreaf | tobiash: heh nice idea - I guess jobname would still risk to cause overlaps though | 12:48 |
| tobiash | job names are unique | 12:49 |
| *** jpena is now known as jpena|lunch | 12:49 | |
| andreaf | tobiash: heh ok you mean using the full job name - yeah that works | 12:50 |
| andreaf | tobiash: also I could have a single dictionary where keys are job names and values are lists and have ansible merge the lists | 12:51 |
| tobiash | andreaf: that would also work | 12:51 |
| tobiash | that's probably better | 12:51 |
| *** _ari_|brno is now known as _ari_ | 12:56 | |
| *** rlandy has joined #zuul | 13:20 | |
| Shrews | pabelanger: clarkb: corvus: So last night I had a thought about this fix for a wedged provider (per https://review.openstack.org/539248). Instead of declining the request based on whether we determine all of our nodes are unsued (which is not easily determined), why don't we choose to delete the oldest unused node if we are about to pause request handling? Then we are guaranteed to unpause "soon", and we don't | 13:27 |
| Shrews | unnecessarily decline requests (which could happen for a long time until a request comes in that uses one of the already available nodes). | 13:27 |
| Shrews | This seems to be a MUCH better solution since it is pro-active rather than reactive. | 13:28 |
| Shrews | If there is no unused node to delete, then resources will free up through normal means anyway. | 13:29 |
| Shrews | So far, I've not been able to see any downside to this. | 13:29 |
| tobiash | Shrews: I think that sounds good | 13:32 |
| Shrews | tobiash: cool. turns out deciding if 1 node is unused is MUCH more easily determined than deciding if ALL nodes are unused. | 13:34 |
| *** cmurphy has joined #zuul | 13:34 | |
| tobiash | and cheaper | 13:34 |
| cmurphy | what is the difference between a playbook using `hosts: all` and `hosts: primary`? I notice that run.yaml playbooks seem to use the first and post.yaml seem to use the second but I don't know why | 13:36 |
| cmurphy | specifically i'm trying to help figure out why this patch to run a new job on an opensuse-423 nodeset isn't running any plays and wondering if that's why https://review.openstack.org/#/c/520063 | 13:42 |
| Shrews | cmurphy: that job seems to be using nodeset opensuse-423, which does not define a 'primary' node, which i think comes from here: https://git.openstack.org/cgit/openstack-infra/openstack-zuul-jobs/tree/zuul.d/nodesets.yaml#n20 | 13:51 |
| *** jpena|lunch is now known as jpena | 13:51 | |
| Shrews | cmurphy: 'hosts: all' is sort of a catch-all | 13:51 |
| cmurphy | Shrews: oh that's interesting | 13:52 |
| cmurphy | Shrews: is there any particular reason we should or shouldn't define a primary for that nodeset? | 13:53 |
| Shrews | cmurphy: *shrug* | 13:53 |
| Shrews | you can see other nodesets define a primary, but they all have additional nodes in the set too | 13:54 |
| Shrews | well, not all i guess | 13:54 |
| cmurphy | if the playbook has hosts: all and the nodeset has multiple nodes would the job get run on all of them? | 13:55 |
| Shrews | yes | 13:55 |
| cmurphy | okay so we probably don't want to change to hosts: all | 13:55 |
| Shrews | in this case, looks like you can use 'all' since there is only a single node | 13:56 |
| Shrews | should be the same result as 'hosts: opensuse-423' | 13:56 |
| Shrews | looks like the use 'primary' was just a copy-pasta error maybe | 13:57 |
| cmurphy | well the playbook was written for a job using nodeset legacy-ubuntu-xenial which has primary defined | 13:58 |
| Shrews | legacy-opensuse-423 had a 'primary' node: https://git.openstack.org/cgit/openstack-infra/openstack-zuul-jobs/tree/zuul.d/nodesets.yaml#n134 | 13:58 |
| Shrews | yeah, looks like all the legacy ones do | 13:59 |
| cmurphy | okay sounds like changing to hosts: all is going to be fine | 13:59 |
| cmurphy | thanks Shrews | 13:59 |
| *** yolanda has quit IRC | 14:01 | |
| *** yolanda has joined #zuul | 14:03 | |
| *** elyezer has quit IRC | 14:06 | |
| *** elyezer has joined #zuul | 14:07 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/nodepool master: Allow operator to set held node's TTL https://review.openstack.org/539493 | 14:11 |
| mordred | andreaf, tobiash: haven't read the whole scrollback, but yes, currently only dicts are merged, lists are replaced | 14:16 |
| andreaf | mordred: cool thanks | 14:17 |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Fix lazy initialization of GithubUser with apps https://review.openstack.org/539545 | 14:17 |
| *** dkranz has joined #zuul | 14:17 | |
| mordred | Shrews: yes, I like your idea | 14:17 |
| Shrews | mordred: bueno. oh, i owe you reviews today too. will try to get to that | 14:19 |
| Shrews | i also owe myself breakfast. biab | 14:21 |
| *** sshnaidm is now known as sshnaidm|mtg | 14:27 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/nodepool master: Allow operator to set held node's TTL https://review.openstack.org/539493 | 14:34 |
| fungi | cmurphy: Shrews: the "primary" node was used to accommodate old-style multinode jobs converted from v2 configuration, but native v3 jobs are no longer constrained to a primary/subnode model | 14:51 |
| *** JasonCL has joined #zuul | 14:51 | |
| fungi | and i'm pretty sur single-node jobs were converted as if they were just a special case "one-node multinode" design | 14:52 |
| fungi | so a primary with no subnodes | 14:52 |
| fungi | mainly because the nodepool metadata used to differ between primary and subnodes | 14:53 |
| *** elyezer has quit IRC | 14:53 | |
| fungi | and nodepool v2 treated stand-alone nodes as a primary with no other subnodes | 14:54 |
| cmurphy | why then did this single node job get converted to hosts: all instead of hosts: primary if it doesn't really matter? http://git.openstack.org/cgit/openstack/keystone/tree/playbooks/legacy/keystone-dsvm-functional/run.yaml | 14:55 |
| cmurphy | i guess maybe there was no way of knowing it was a single node job when it was converted | 14:55 |
| *** elyezer has joined #zuul | 14:56 | |
| Shrews | cmurphy: yeah, the script we used to convert them wasn't perfect | 15:00 |
| * Shrews still gets chills thinking about that script.... *shudder* | 15:01 | |
| pabelanger | also using hosts: all means we don't need to add specific nodesets to jobs. A little easier to switch between fedora / ubuntu / etc | 15:03 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/nodepool master: Allow operator to set held node's TTL https://review.openstack.org/539493 | 15:10 |
| *** elyezer has quit IRC | 15:11 | |
| *** elyezer has joined #zuul | 15:24 | |
| corvus | andreaf: well, a lot of the devstack stuff works by having a 'fake' value of "True", so it may not be that odd. | 15:26 |
| *** Guest87240 is now known as mgagne | 15:27 | |
| *** mgagne has joined #zuul | 15:27 | |
| andreaf | corvus: heh you're not jeblair anymore - that's why I didn't see you online :) | 15:28 |
| mordred | Shrews: never think about that script again | 15:31 |
| andreaf | corvus: well if I have a dict with true / false I kind of set the expectation that false actually means false and something other than False / True triggers an error | 15:31 |
| andreaf | corvus: but actually it may even be easier to implement that than the dict of lists | 15:32 |
| corvus | andreaf: yeah, the devstack conf modules also honor false -- they act as if the value wasn't there. that lets child jobs override parents and remove things | 15:32 |
| *** sshnaidm|mtg has quit IRC | 15:32 | |
| *** ChanServ changes topic to "Discussion of the project gating system Zuul | Docs: http://docs.openstack.org/infra/zuul/ | Source: https://git.openstack.org/cgit/openstack-infra/zuul/ | Roadmap: https://storyboard.openstack.org/#!/board/53 | Channel logs: http://eavesdrop.openstack.org/irclogs/%23zuul/" | 15:33 | |
| andreaf | corvus: the thing I'm struggling with atm is providing some kind of backward compatibility - if the var is a dict to A, if it's a list to B - how do I do that check in ansible? | 15:33 |
| corvus | tristanC: good idea, topic changed, thanks :) | 15:34 |
| *** sshnaidm|mtg has joined #zuul | 15:34 | |
| corvus | andreaf: i bet you could use a jinja expression for that. something like "when: varname is mapping" maybe? | 15:36 |
| Shrews | FYI, I have no water at home (water main break). I may be in and out as I move to different places to remain comfortable today. | 15:44 |
| mordred | Shrews: wait - you had running water to begin with? | 15:47 |
| clarkb | Shrews: yes I think deleting oldest ready node if there is a ready node would work | 15:47 |
| clarkb | then we only delete one node instead of potentially 50 for example | 15:47 |
| andreaf | corvus: when extensions_to_txt | type_debug == 'dict' works fine | 15:51 |
| andreaf | corvus: I'm not sure type_debug was meant for that though | 15:51 |
| *** sshnaidm|mtg has quit IRC | 15:52 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/nodepool master: Allow operator to set held node's TTL https://review.openstack.org/539493 | 15:52 |
| clarkb | looks like mirror02.dfw.rax.openstack.org has the dfw mirror cname and cacti makes it looks reasonably happy? | 15:56 |
| clarkb | er ww | 15:56 |
| *** sshnaidm|mtg has joined #zuul | 16:08 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/nodepool master: Allow operator to set held node's TTL https://review.openstack.org/539493 | 16:21 |
| tobiash | Shrews: do you think it's possible to add a 'fast track' for requests which can be fulfilled immediately by ready nodes regardless if the provider is paused or not? | 16:44 |
| tobiash | that would be also an optimization if we're at quota, provider is paused but there are ready nodes of some types | 16:45 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: [WIP] zuul autohold: allow operator to specify nodes TTL https://review.openstack.org/539596 | 16:48 |
| Shrews | tobiash: the question should be "is that a good idea" since it could allow requests to jump our priority queue | 16:49 |
| Shrews | I don't have the answer to that | 16:49 |
| clarkb | I think if we have excess ready nodes able to jump the queue that is fine (at least thinking about it really quickly that allows us to free resources for older requests) | 16:50 |
| tobiash | that's also what I thought | 16:52 |
| tobiash | as long as we don't create a starvation problem | 16:52 |
| *** sshnaidm|mtg is now known as sshnaidm | 17:01 | |
| tobiash | pabelanger: did you see SpamapS's comment on https://review.openstack.org/#/c/537953/ ? | 17:01 |
| pabelanger | tobiash: yah, I can push up a quick fix. | 17:02 |
| tobiash | kk | 17:02 |
| *** harlowja has joined #zuul | 17:02 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Enabled ssh retries for ansible https://review.openstack.org/537953 | 17:05 |
| pabelanger | okay, updated | 17:05 |
| tobiash | :) | 17:06 |
| tobiash | mordred, ianw: I think https://storyboard.openstack.org/#!/story/2001487 can be resolved now right? | 17:11 |
| *** bhavik1 has joined #zuul | 17:12 | |
| mordred | tobiash: yah | 17:16 |
| mordred | tobiash: done | 17:16 |
| *** leifmadsen_ is now known as leifmadsen | 17:21 | |
| *** myoung is now known as myoung|rabbit | 17:23 | |
| *** bhavik1 has quit IRC | 17:34 | |
| jlk | So, GitHub just shipped a feature that officially recognizes the Co-Authored-By lines in a commit, and gives author credit in the UI to those listed. | 17:40 |
| jlk | e.g. https://github.com/openstack-infra/zuul/commit/4fc12549072d02b97fa46ab36b4618f320ce507f | 17:40 |
| jlk | https://github.com/blog/2496-commit-together-with-co-authors | 17:42 |
| tobiash | jlk: cool | 17:51 |
| tobiash | jlk: do you know if there's an api to query just the code review status of a pr? | 17:51 |
| tobiash | I didn't find one while trying to evaluate needed reviews for the gate in zuul | 17:52 |
| jlk | toabctl: one sec. | 17:52 |
| jlk | toabctl: so, you kind of have to sort that out for yourself. GET /repos/:owner/:repo/pulls/:number/reviews You can get all the reviews for a given PR, and then you have to determine A) what was the last review that mattered per person, and does that person have merge rights. | 17:54 |
| jlk | there isn't an exposed API of "this PR has all the necessary reviews to be merged", that's more about branch protection settings, which include other factors. | 17:54 |
| mordred | jlk: neat! | 17:56 |
| tobiash | jlk: hrm, we probably need this in the future | 17:59 |
| tobiash | When working with codeowners this is going to be complicated | 17:59 |
| jlk | yeah, it's a tough nut to crack | 18:01 |
| tobiash | Maybe I should open a support ticket with a feature request | 18:01 |
| jlk | I think what we want to know is "are all requested reviews met" | 18:01 |
| tobiash | Yes | 18:01 |
| jlk | that's part of what you get with code owners. | 18:01 |
| clarkb | corvus added a similar feature to gerrit | 18:01 |
| jlk | You can get https://developer.github.com/v3/pulls/review_requests/#list-review-requests | 18:01 |
| jlk | which is a list of all the requested reviews, and map them up to what you've figured out from the existing reviews | 18:02 |
| jlk | In theory, figuring this all out would be easier with GraphQL, at least in the number of queries necessary. But that's still not on for Apps | 18:03 |
| tobiash | Ah, ok, that could make a workaround possible | 18:03 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul-jobs master: Look for 0 in .stestr directory instead of failing https://review.openstack.org/539624 | 18:11 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul-jobs master: Skip checksums on stat calls in fetch-subunit-output https://review.openstack.org/539625 | 18:11 |
| clarkb | mordred: ^ changes like that make me wonder if ansible should have a checksum module for checksumming and stat should just stat | 18:12 |
| *** jpena is now known as jpena|away | 18:13 | |
| mordred | clarkb: yes. I very much do not like that stat checksums by default | 18:13 |
| mordred | clarkb: to be fair - what I REALLY want most of the time is an 'exists' module that just returns a single bool | 18:14 |
| mordred | clarkb: rather than needing to look at stat_results.stat.exists | 18:14 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: [WIP] zuul autohold: allow operator to specify nodes TTL https://review.openstack.org/539596 | 18:15 |
| jlk | mordred: I think a lot of people want that too | 18:17 |
| jlk | "does this exist yes/no" | 18:17 |
| mordred | jlk: I'm writing it right now | 18:17 |
| mordred | jlk: I have reached my annoyance threshold | 18:17 |
| jlk | developing in anger | 18:18 |
| mordred | it's how I work :) | 18:18 |
| *** myoung|rabbit is now known as myoung | 18:22 | |
| *** dkranz has quit IRC | 18:44 | |
| mordred | jlk: https://github.com/ansible/ansible/pull/35572 | 18:45 |
| mordred | jlk: I tagged you in it :) | 18:45 |
| jlk | reading | 18:46 |
| corvus | Shrews, mhu: should we drop the 'nodepool hold' command? (i think i thought we had already). in v3, it can really only be used to hold a ready node, not a used node. is that still useful? | 18:50 |
| corvus | prompted by https://review.openstack.org/539493 | 18:51 |
| *** harlowja has quit IRC | 18:52 | |
| corvus | Shrews: also https://review.openstack.org/535553 and parent are ready for your +3 | 18:52 |
| corvus | mordred: ^ you may be interested | 18:52 |
| *** sshnaidm is now known as sshnaidm|off | 18:55 | |
| mordred | corvus: ooh, that's exciting | 18:56 |
| jlk | mordred: I've thrown some words back at you on that PR. | 19:00 |
| mordred | woot | 19:06 |
| mordred | jlk: done | 19:20 |
| jlk | mordred: have you thought about making it a fact module, to skip the "register" part, and have it set a fact name based on the path? (that might be TOO magic) | 19:21 |
| jlk | I don't know how long it's been doing this, but I like that GitHub UI shows when a force push happens in a PR timeline. | 19:23 |
| *** jpena|away is now known as jpena|off | 19:24 | |
| corvus | i've culled through many of the open nodepool changes; we're down to 52 open changes now, a bit less than half are pre-v3 changes. it'd be great of other zuul-core folks can pitch in on abandoning or WIPing old changes. here's the messages i've been leaving: http://paste.openstack.org/show/658218/ | 19:26 |
| *** dkranz has joined #zuul | 19:27 | |
| corvus | mordred: about half of the nodepool ones left are from you, so you may be particularly suited to triaging those :) | 19:27 |
| *** dkranz has quit IRC | 19:29 | |
| *** harlowja has joined #zuul | 19:37 | |
| Shrews | corvus: huh, i thought we had removed the hold command | 19:39 |
| Shrews | it's pretty worthless right now | 19:40 |
| corvus | Shrews: apparently a shared delusion! | 19:42 |
| Shrews | those are the best | 19:43 |
| Shrews | corvus: i left a comment on the parent change | 19:47 |
| mordred | corvus: I think most of my nodepool patches in that stack are bong - but I'll go through them | 19:53 |
| mordred | jlk: yah - the force push visualization is very helpful | 19:54 |
| mordred | jlk: for the other thing - I think *I* would like that, but I have a hunch it's too magic to get right | 19:54 |
| jlk | nod | 19:54 |
| mordred | jlk: what I REALLY want, lf course, is 'when: exists(path)' | 19:54 |
| tobiash | like a remote lookup plugin? | 19:55 |
| tobiash | is that possible? | 19:55 |
| openstackgerrit | Merged openstack-infra/zuul master: Enabled ssh retries for ansible https://review.openstack.org/537953 | 19:55 |
| mordred | tobiash: nope | 19:55 |
| mordred | tobiash: I'm just wishing for magical ponies | 19:56 |
| mordred | jlk: zomg. we could do fact setting and go full-on old school unix on it - have it set a "last_exists" fact every time - kind of like $? | 19:56 |
| jlk | erm.... | 19:56 |
| mordred | exists: path=foo .... when: last_exists ... | 19:56 |
| mordred | **terrible** idea - just saying | 19:56 |
| *** elyezer has quit IRC | 20:14 | |
| *** dkranz has joined #zuul | 20:14 | |
| openstackgerrit | Andrea Frittoli proposed openstack-infra/zuul-jobs master: Change the list of extensions to a dict https://review.openstack.org/539683 | 20:57 |
| openstackgerrit | Andrea Frittoli proposed openstack-infra/zuul-jobs master: Change the list of extensions to a dict https://review.openstack.org/539683 | 20:58 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Merger: retry network operations https://review.openstack.org/539356 | 21:20 |
| corvus | Shrews: the func tests passed the recheck so i approved 535553 | 21:22 |
| Shrews | corvus: great thx | 21:23 |
| *** dkranz has quit IRC | 21:58 | |
| *** myoung is now known as myoung|off | 22:09 | |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Support the fragment form of Gerrit URLs https://review.openstack.org/539705 | 22:13 |
| corvus | mordred, Shrews: ^ | 22:14 |
| corvus | clarkb: can you review https://review.openstack.org/538353 when you get a chance? | 22:16 |
| clarkb | jlk: ya looking | 22:22 |
| clarkb | er | 22:22 |
| clarkb | corvus: ^ | 22:22 |
| * jlk waves | 22:22 | |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Add available RAM to statsd https://review.openstack.org/539707 | 22:22 |
| corvus | jhesketh: with ^ are you okay to approve https://review.openstack.org/539426 ? | 22:23 |
| corvus | jlk: https://review.openstack.org/539545 seems to me like an easy github review | 22:26 |
| * jlk reads | 22:27 | |
| corvus | tobiash: https://review.openstack.org/537428 approved with comments | 22:48 |
| openstackgerrit | Merged openstack-infra/zuul master: Fix statsd documentation about events https://review.openstack.org/539482 | 22:53 |
| corvus | SpamapS: what's the latest on re2? | 22:57 |
| openstackgerrit | Merged openstack-infra/zuul master: Fix lazy initialization of GithubUser with apps https://review.openstack.org/539545 | 23:00 |
| SpamapS | corvus: I have not gotten a reply back on my latest PR push. I'll ping the reviewers. | 23:03 |
| corvus | SpamapS: ack | 23:04 |
| SpamapS | https://github.com/facebook/pyre2/pull/10 | 23:04 |
| SpamapS | just commented | 23:04 |
| corvus | SpamapS: should 536389 Depends-On that? | 23:05 |
| corvus | SpamapS: (i mean, zuul won't do anything at this point since it's not a defined project, but that's not the only reason to include that) | 23:06 |
| corvus | i guess it really depends on that plus a release | 23:06 |
| SpamapS | corvus: yes actually. | 23:06 |
| SpamapS | We could ask if they want to join our cross-repo testing party. :) | 23:07 |
| corvus | maybe someday -- though i also think we can get to the point where you could depends-on that PR and actually have it work, without otherwise adding their repo to the system. | 23:08 |
| corvus | (basically, dynamically evaluate required-projects. which is also one of my ideas for reducing memory usage, so that may come sooner than you would otherwise think) | 23:08 |
| SpamapS | True, I think the way the driver works now it does a search for the PR's status, so while it wouldn't trigger gate entry for the zuul patch on commit of the pyre2 commit, a Depends-On would prevent it from merging to zuul until the pyre2 PR merges, yes? | 23:09 |
| corvus | SpamapS: i think so, though i'm not 100% sure that works for projects that aren't in the tenant. (though there's no reason it shouldn't/couldn't) | 23:11 |
| SpamapS | further experimentation is necessary | 23:11 |
| corvus | ++ | 23:11 |
| *** rlandy is now known as rlandy|bbl | 23:35 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!