| *** jamesmcarthur has joined #zuul | 00:03 | |
| *** jamesmcarthur has quit IRC | 00:08 | |
| *** jamesmcarthur has joined #zuul | 00:12 | |
| *** jamesmcarthur has quit IRC | 00:16 | |
| pabelanger | ianw: so, how can I have DIB do a noop on an image? for example | 01:17 |
|---|---|---|
| pabelanger | ELEMENTS_PATH=`pwd` /opt/venv/nodepool-3.5.0/bin/disk-image-create -x -n --no-tmpfs -o test.qcow2 vyos | 01:17 |
| pabelanger | vyos elements contain operating-system and no-final-image | 01:17 |
| pabelanger | however, DIB errors out | 01:18 |
| pabelanger | http://paste.openstack.org/show/748979/ | 01:18 |
| ianw | pabelanger: ummm, you don't want a complete no-op like the fake builder? | 01:18 |
| pabelanger | right | 01:18 |
| pabelanger | I don't want to do anything with the existing image, just have DIB rename it to the value that nodpeool-builder passes | 01:18 |
| pabelanger | because this is a network vendor image | 01:19 |
| pabelanger | I'm using cloud-images now, but really want diskimages, so nodepool can handle the uploads for me | 01:19 |
| ianw | ahh, yeah i was going to say that sounds like a case for a pre-existing image | 01:19 |
| ianw | can you run that with -x ? but i don't think this case has ever been considered at all from dib's pov | 01:20 |
| pabelanger | yah, 1 sec | 01:20 |
| pabelanger | https://pastebin.com/raw/sVXjSKqx | 01:21 |
| pabelanger | So, I'm thinking 2 options, 1 is to expose builder command into diskimages section, so I can replace diskimage-builder with a fake sh script, this is how our testing works. | 01:21 |
| pabelanger | Or maybe we add a cloudimages section, along side diskimages, which just does the upload part | 01:22 |
| ianw | yeah, you're into https://git.openstack.org/cgit/openstack/diskimage-builder/tree/diskimage_builder/lib/common-functions#n340 which is assuming we've mounted and are in a chroot | 01:24 |
| ianw | i feel like the cloudimages idea sounds very reasonable. basically you have a .qcow2 from somewhere else and nodepool manages it from there | 01:26 |
| pabelanger | ianw: cool, thanks. I'll bounce the idea off Shrews, otherwise I think adding support for packer is something I am going to do. The vendor image is build using that, then I could have nodepool-builder do it today. | 01:27 |
| pabelanger | however, zuul-job, SSH to nodepool-builder is how I am planning on doing it right now | 01:28 |
| ianw | yeah different building backends seems useful ... | 01:30 |
| SpamapS | Just allowing the command and args to be configured would be ideal. most builders aren't as compositional as dib.. so it should all fit into the current abstraction. | 01:33 |
| pabelanger | Agree | 01:34 |
| pabelanger | Nice! | 01:36 |
| pabelanger | https://logs.zuul.ansible.com/25/25/dc2cad281a0f75068db0db74a956aa5ae92bf81e/check/vyos-test/fcde131/ | 01:37 |
| pabelanger | that is a vyos node! | 01:37 |
| pabelanger | mind you, it doesn't have python27, only python26 | 01:37 |
| pabelanger | but nodepool launched it | 01:37 |
| pabelanger | Yay | 01:37 |
| *** irclogbot_3 has quit IRC | 01:44 | |
| pabelanger | okay, neat. Using job.host-vars I can set ansible_python_interpreter | 01:46 |
| pabelanger | it would be kinda neat, if I could do that under nodesets directly... but for now, what I got :D | 01:46 |
| pabelanger | I'll deal with rest tomorrow | 01:47 |
| *** edmondsw_ has quit IRC | 01:48 | |
| *** jamesmcarthur has joined #zuul | 01:50 | |
| clarkb | wouldnt it be better to uplosd the image yourself and tell nodepool about it if the image isnt being updated? | 02:02 |
| clarkb | otherwise nodepool will reupload it everyday | 02:02 |
| clarkb | and telling nodepool about a new image isnt much different than telling a cloud? | 02:03 |
| ianw | clarkb: yeah, i mean i was kind of assuming you were pulling in an image frequently, and wanted nodepool to manage just keeping the latest versions | 02:14 |
| clarkb | ya I guess dib could pull from a known location that updates and act as ashuttle | 02:15 |
| ianw | yeah, that was pabelanger's initial point i guess; but dib has a bunch of checks that are assuming it's got a chroot with the image open | 02:16 |
| ianw | i mean, we could just have a DIB_DO_NOTHING=1 flag and basically turn it into a cp ... but it seems a bit odd | 02:16 |
| clarkb | or justrun no elements and let dib chroot then make a newfile? | 02:16 |
| clarkb | its less efficient but should work? | 02:17 |
| clarkb | kindalike the ubuntu element or centos element? justdont run anything more? | 02:17 |
| ianw | yeah, but if you're opening the incoming qcow2, making it into a chroot and the repacking it, and outputting what you hope is the same .qcow2 it seems ... a little crazy. | 02:19 |
| clarkb | ya you may end up with a different filesystem or other unexpected changes too | 02:22 |
| *** bhavikdbavishi has joined #zuul | 02:23 | |
| *** jamesmcarthur has quit IRC | 03:03 | |
| *** raukadah is now known as chandankumar | 03:14 | |
| *** bhavikdbavishi has quit IRC | 03:47 | |
| *** bhavikdbavishi has joined #zuul | 03:47 | |
| *** bjackman has joined #zuul | 03:56 | |
| *** bhavikdbavishi has quit IRC | 04:13 | |
| *** pcaruana has joined #zuul | 04:49 | |
| *** pcaruana has quit IRC | 04:55 | |
| *** bhavikdbavishi has joined #zuul | 04:58 | |
| *** bhavikdbavishi1 has joined #zuul | 04:58 | |
| *** bhavikdbavishi has quit IRC | 05:02 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 05:02 | |
| *** jamesmcarthur has joined #zuul | 05:05 | |
| *** jamesmcarthur has quit IRC | 05:06 | |
| *** jamesmcarthur has joined #zuul | 05:06 | |
| *** jamesmcarthur has quit IRC | 05:10 | |
| *** snapiri has joined #zuul | 05:14 | |
| *** quiquell|off is now known as quiquell|rover | 06:05 | |
| *** AJaeger has quit IRC | 06:06 | |
| bjackman | When Zuul runs ansible, ~/.ansible/tmp gets created in the zuul user's home directory on the test node (i.e. not on the executor itself but on the remote host of the ansible run). Anyone know why that might be? | 06:10 |
| *** AJaeger has joined #zuul | 06:10 | |
| bjackman | The only reference I can find to that path is that it's the default for the local_tmp ansible param. But that is a path on the local host (executor) not the test node IIUC. Also it is always overridden by the ansible.cfg created by the executor server. | 06:11 |
| bjackman | Ah, sorry, scratch that, there's "remote_tmp" as well | 06:15 |
| *** quiquell|rover is now known as quique|rover|brb | 06:22 | |
| *** snapiri has quit IRC | 06:28 | |
| *** snapiri has joined #zuul | 06:29 | |
| *** pcaruana has joined #zuul | 06:30 | |
| *** sshnaidm|off is now known as sshnaidm | 06:53 | |
| *** themroc has joined #zuul | 06:59 | |
| *** quique|rover|brb is now known as quiquell|rover | 07:01 | |
| *** gtema has joined #zuul | 07:09 | |
| *** gtema has quit IRC | 07:33 | |
| *** gtema has joined #zuul | 07:44 | |
| *** jangutter has joined #zuul | 07:49 | |
| *** yolanda_ has joined #zuul | 07:52 | |
| *** jpena|off is now known as jpena | 07:54 | |
| *** bjackman_ has joined #zuul | 08:11 | |
| *** yolanda_ has quit IRC | 08:11 | |
| *** yolanda_ has joined #zuul | 08:11 | |
| *** bjackman has quit IRC | 08:14 | |
| sshnaidm | I have an error: "Unexpected templating type error occurred on ({{ zuul.items|list }}): 'builtin_function_or_method' object is not iterable" - is zuul.items a list? How do I get its "items"? | 08:27 |
| *** badboy has quit IRC | 08:30 | |
| *** hashar has joined #zuul | 08:49 | |
| *** yolanda_ has quit IRC | 09:09 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: Web: plug the authorization engine https://review.openstack.org/640884 | 09:19 |
| *** tobberydberg has joined #zuul | 09:29 | |
| sshnaidm | tristanC, mhu any ideas? ^^ | 09:31 |
| tristanC | sshnaidm: iiuc, zuul.items is already a list, so you don't have to use the |list filter | 09:35 |
| mhu | sshnaidm, is that what you're looking for? https://zuul-ci.org/docs/zuul/user/jobs.html?highlight=variables#var-zuul | 09:35 |
| mhu | yep, more precisely: https://zuul-ci.org/docs/zuul/user/jobs.html?highlight=variables#var-zuul.items | 09:35 |
| mhu | tristanC is right | 09:36 |
| sshnaidm | mhu, tristanC when I do "debug: var=zuul.items " it shows me "<built-in method items of dict object at 0x7f68ed28da08>" | 09:42 |
| tristanC | sshnaidm: that's odd, where is that happening? | 09:50 |
| sshnaidm | tristanC, http://logs.rdoproject.org/09/20009/8/check/tripleo-ci-reproducer-fedora-28-libvirt-standalone-dryrun/396111c/job-output.txt.gz | 09:55 |
| tristanC | sshnaidm: it seems like rejectattr produces a generator | 10:06 |
| *** yolanda_ has joined #zuul | 10:11 | |
| tristanC | sshnaidm: it seems like you could use json_query instead to filter zuul.items | 10:12 |
| tristanC | sshnaidm: oh I know, zuul.items get's interpreted as the "items" function of zuul, you should try zuul["items"] instead | 10:16 |
| sshnaidm | tristanC, ack | 10:16 |
| *** yolanda_ has quit IRC | 10:38 | |
| *** electrofelix has joined #zuul | 10:48 | |
| *** yolanda_ has joined #zuul | 10:48 | |
| quiquell|rover | tristanC: A Depends-On to Depends-On: gnocchixyz/gnocchi/pull/995 is correct ? | 10:54 |
| quiquell|rover | tristanC: it's a configured github project | 10:54 |
| quiquell|rover | tristanC: or do I have to preprend github.com ? | 10:54 |
| tristanC | quiquell|rover: the new Depends-On syntax needs complete url | 10:54 |
| quiquell|rover | tristanC: ack so https://github.com/gnocchixyz/gnocchi/pull/995 | 10:55 |
| tristanC | quiquell|rover: yes | 10:55 |
| quiquell|rover | tristanC: thanks | 10:55 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: Web: plug the authorization engine https://review.openstack.org/640884 | 11:02 |
| *** bhavikdbavishi has quit IRC | 11:02 | |
| *** hashar has quit IRC | 11:18 | |
| *** hashar has joined #zuul | 11:19 | |
| *** jpena is now known as jpena|lunch | 11:27 | |
| *** rfolco has joined #zuul | 11:33 | |
| *** bhavikdbavishi has joined #zuul | 11:43 | |
| *** pcaruana has quit IRC | 11:47 | |
| pabelanger | ianw: clarkb: Yah, I wouldn't want DIB to modify the qcow2 image, until I had something in the actually install.d phase | 11:56 |
| *** rlandy has joined #zuul | 11:57 | |
| ianw | pabelanger: oh, are you likely to *have* something in the install phase? | 11:57 |
| *** quiquell|rover is now known as quique|rover|eat | 11:57 | |
| ianw | i was really imaging a situation where the image came from a remote daily build or similar, which would bake in any modifications there | 11:58 |
| *** rlandy is now known as rlandy|ruck | 12:02 | |
| *** edmondsw has joined #zuul | 12:04 | |
| *** jamesmcarthur has joined #zuul | 12:17 | |
| *** jpena|lunch is now known as jpena | 12:22 | |
| pabelanger | ianw: no, I don't expect install phase for this image | 12:29 |
| *** jamesmcarthur has quit IRC | 12:30 | |
| *** pcaruana has joined #zuul | 12:38 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: Web: plug the authorization engine https://review.openstack.org/640884 | 12:39 |
| *** quique|rover|eat is now known as quiquell|rover | 12:40 | |
| *** jamesmcarthur has joined #zuul | 12:48 | |
| *** gtema has quit IRC | 12:49 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul-jobs master: Add zuul_debug_info_enabled role setting https://review.openstack.org/650880 | 12:53 |
| *** Shrews has quit IRC | 12:56 | |
| *** quiquell|rover is now known as quique|rover|bbl | 12:57 | |
| *** gtema has joined #zuul | 12:57 | |
| *** Shrews has joined #zuul | 12:58 | |
| *** bjackman_ has quit IRC | 13:01 | |
| *** pwhalen has joined #zuul | 13:03 | |
| pabelanger | zuul-jobs-core: I'd love a review on https://review.openstack.org/650880/ | 13:21 |
| pabelanger | adds a toggle to zuul_debug_info task | 13:21 |
| tristanC | pabelanger: speaking of toggle, could you please re-review https://review.openstack.org/632620 | 13:28 |
| AJaeger | pabelanger: left a suggestion for the README file | 13:28 |
| *** gtema has quit IRC | 13:29 | |
| *** bhavikdbavishi has quit IRC | 13:30 | |
| *** pcaruana has quit IRC | 13:31 | |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: Zuul Web: add /api/user/authorizations endpoint https://review.openstack.org/641099 | 13:34 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: authentication config: add optional token_expiry https://review.openstack.org/642408 | 13:34 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul-jobs master: Add zuul_debug_info_enabled role setting https://review.openstack.org/650880 | 13:35 |
| pabelanger | AJaeger: fixed | 13:36 |
| *** pcaruana has joined #zuul | 13:36 | |
| pabelanger | tristanC: looking | 13:36 |
| pabelanger | tristanC: +1, left comment about adding a warning | 13:40 |
| pabelanger | I'll defer to others to approve | 13:40 |
| *** quique|rover|bbl is now known as quiquell|rover | 13:51 | |
| fungi | osf board of directors meeting is underway: https://wiki.openstack.org/wiki/Governance/Foundation/8April2019BoardMeeting | 14:12 |
| fungi | kata containers presentation for potential oip confirmation is going on now, zuul's should be starting shortly | 14:13 |
| pabelanger | fungi: do you happen to know dialin info? | 14:13 |
| clarkb | pabelanger: let me find a link for you | 14:14 |
| pabelanger | link to zoom doesn't seem to show it | 14:14 |
| pabelanger | (without downloading something first) | 14:14 |
| pabelanger | clarkb: thanks | 14:14 |
| fungi | dial in numbers are available: https://zoom.us/zoomconference | 14:14 |
| clarkb | https://wiki.openstack.org/wiki/Governance/Foundation/8April2019BoardMeeting has the zoom "room" number and ^ has the POTS numbers | 14:14 |
| pabelanger | thanks | 14:15 |
| clarkb | (we should really put the POTS contact info on the wiki) | 14:15 |
| fungi | the conference id is the string of numbers at the end of the zoom url, so 9574980705 | 14:15 |
| *** themroc has quit IRC | 14:15 | |
| * fungi is dialled in by phone, for the record, doesn't trust the zoom client/browser plug-in | 14:15 | |
| clarkb | fungi: worried another excavator will find that fiber line? | 14:16 |
| pabelanger | okay, connected. Thanks | 14:16 |
| fungi | you never kjnow | 14:16 |
| fungi | but no, by "trust" i mean i am unwilling to let it execute on my computers | 14:17 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: Zuul Web: add /api/user/authorizations endpoint https://review.openstack.org/641099 | 14:19 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: authentication config: add optional token_expiry https://review.openstack.org/642408 | 14:19 |
| pabelanger | zuul-jobs-core: I'd also like a review on https://review.openstack.org/648773/ to expose ara_executable path, which is an additional change on top of our recent zuul-executor venv path fixes | 14:30 |
| clarkb | pabelanger: do we still need that if the venv is exposed? | 14:32 |
| clarkb | (maybe we need a doc update that says 'you can put python tools into the virtualenv and they will be available to the executor') | 14:32 |
| dmsimard | pabelanger: I'm curious about the use case | 14:32 |
| dmsimard | is ara not installed in the expected location ? | 14:33 |
| pabelanger | clarkb: right, once we release zuul, it will be fixed. However, I'm currently blocked on a zuul release, which I am unsure when the next one will be. This is my way to get it working now | 14:33 |
| pabelanger | dmsimard: see https://review.openstack.org/649427 | 14:34 |
| pabelanger | clarkb: dmsimard: however, that doesn't seem to fix ara getting loaded by zuul-executor, as not in path. So that won't 100% fix my issue, since databae isn't created | 14:37 |
| dmsimard | pabelanger: I guess this doesn't work: https://opendev.org/openstack-infra/zuul/src/branch/master/zuul/executor/server.py#L40-L43 | 14:41 |
| dmsimard | if ara is installed in the venv and not for the zuul executor proper | 14:42 |
| dmsimard | I guess you should be able to tell if the executor is able to find the callback by looking at the generated ansible.cfg file https://opendev.org/openstack-infra/zuul/src/branch/master/zuul/executor/server.py#L1748-L1766 | 14:43 |
| pabelanger | dmsimard: right, doing a test now | 14:45 |
| pabelanger | on 3.7.1, I don't have ara loaded as callback | 14:45 |
| pabelanger | want to make sure 3.8.0 has is | 14:46 |
| *** dkehn has joined #zuul | 14:52 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Assert ara database is loaded from ansible venv https://review.openstack.org/650917 | 14:58 |
| pabelanger | dmsimard: ^confirmed to be working | 14:59 |
| pabelanger | tobiash_: ^ | 15:00 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: Zuul Web: add /api/user/authorizations endpoint https://review.openstack.org/641099 | 15:07 |
| Shrews | I need someone to merge this nodepool change plz https://review.openstack.org/650385 | 15:13 |
| clarkb | Shrews: I can look when monty is done talking | 15:14 |
| Shrews | k, thx | 15:15 |
| *** bhavikdbavishi has joined #zuul | 15:15 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: [WIP] Set focus on xtem.js to terminal https://review.openstack.org/650275 | 15:20 |
| openstackgerrit | Matthieu Huin proposed openstack-infra/zuul master: authentication config: add optional token_expiry https://review.openstack.org/642408 | 15:28 |
| *** rfolco is now known as rfolco|lunch | 15:29 | |
| fungi | monty holds a semaphore lock on our conscious thought processes | 15:32 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Set focus on xtem.js to terminal https://review.openstack.org/650275 | 15:32 |
| pabelanger | tobiash_: ^should set focus() properly now | 15:32 |
| * mordred releases the lock | 15:44 | |
| *** quiquell|rover is now known as quiquell|off | 15:50 | |
| *** rfolco|lunch is now known as rfolco | 15:58 | |
| *** rlandy|ruck is now known as rlandy | 16:03 | |
| Shrews | him, i don't think our new registry jobs are very ipv6 friendly: http://logs.openstack.org/85/650385/3/gate/nodepool-upload-image/c7a0c31/job-output.txt.gz#_2019-04-08_15_54_03_022549 | 16:06 |
| Shrews | s/him/hrm/ curse autocorrect | 16:06 |
| mordred | Shrews: yes - there is an ipv6 issue with the registry jobs on the ipv6-only clouds - there is a patch up from corvus but he didn't want to land it and then go on vacation | 16:12 |
| clarkb | we should land that now and watch it | 16:13 |
| mordred | oh - it landed | 16:13 |
| mordred | https://review.openstack.org/#/c/650473/ | 16:13 |
| clarkb | ok cool so we may have more debugging to do | 16:13 |
| clarkb | that actually looks like a docker bug | 16:13 |
| mordred | are the nodepool jobs using the roles? ah - awesome | 16:13 |
| clarkb | we may be the first people to try and ipv6 with docker | 16:13 |
| mordred | heh | 16:13 |
| clarkb | if you look at the rror it escaped the ipv6 addr so that it could :5000 and have a proper "url" | 16:14 |
| clarkb | except docker doesn't do real urls and I'm sure that means they aren't parsing the ipv6 properly | 16:14 |
| clarkb | is there some other method of escpae? | 16:14 |
| clarkb | fungi: ^ you may know | 16:14 |
| fungi | [] | 16:15 |
| clarkb | fungi: ya that is what we've done "[2607:ff68:100:54:f816:3eff:fe7c:e98a]:5000/zuul/nodepool:latest" is not a valid repository/tag: invalid reference format | 16:15 |
| fungi | ahh, that's the traditional method of escaping for an address+port url | 16:16 |
| clarkb | yup | 16:16 |
| mordred | stellar | 16:16 |
| clarkb | makes me think docker is broken | 16:16 |
| fungi | where "broken" might mean they were unaware of convention and invented their own syntax | 16:17 |
| mordred | or quite simply haven't had anyone attempt to push an image to a local ipv6-only registry that doesn't have a hostname | 16:17 |
| fungi | yes, also entirely possible | 16:18 |
| logan- | /etc/hosts to the rescue | 16:18 |
| clarkb | logan-: ya that may be our best option | 16:18 |
| clarkb | as well as filing a bug with docker | 16:18 |
| fungi | the technical term is an "address literal" | 16:19 |
| fungi | ietf rfc 2732 | 16:21 |
| *** manjeets has joined #zuul | 16:22 | |
| *** bhavikdbavishi has quit IRC | 16:25 | |
| pabelanger | with PTL out this week, is it safe to say no zuul release this week? | 16:25 |
| mhu | pabelanger, when does it start? | 16:26 |
| fungi | probably better to defer, though if there's something which needs an urgent release to fix then i wouldn't say it's definite | 16:26 |
| pabelanger | mhu: PTL out? IIRC, vacation started Saturday | 16:27 |
| fungi | remember, our project leader has no additional rights over other zuul maintainers, only additional responsibilities. if we need a release we can make one happen even when he's taking a much-needed break | 16:27 |
| mhu | thx | 16:27 |
| mhu | oh, I misread, I thought you meant the Denver Summit | 16:28 |
| pabelanger | fungi: not right now, as I have to run to appointment, but maybe we could discuss the idea of beta / RC releases, up to pypi. Most of the tooling today for us depends on getting zuul from pypi, not that it could be changed to consume from git, but really trying to run the released versions of zuul / nodepool. | 16:29 |
| fungi | the denver summit starts monday april 29 | 16:29 |
| fungi | pabelanger: cool, that sounds like a less urgent thing which we shouldn't decide to start doing on very little notice anyway | 16:30 |
| pabelanger | fungi: agree, not this week for sure | 16:30 |
| pabelanger | okay, afk for a bit | 16:31 |
| fungi | i would advise starting an ml thread about it | 16:31 |
| pabelanger | +1 | 16:31 |
| fungi | when you have time | 16:31 |
| clarkb | fungi: pabelanger note that the general process has been to update openstack-infra/opendev's zuul to the current commit and run it for a day or two to check there aren't any major issues | 16:37 |
| clarkb | openstack-infra/opendev is currently running 3.7.0 | 16:38 |
| clarkb | (says zuul status page) | 16:38 |
| clarkb | and with the openstack release this week we probably don't want to update zuul today through thursday | 16:38 |
| clarkb | so ya I don't expect a release this week if we want to stick to that burn in process | 16:38 |
| fungi | well, at least through wednesday | 16:43 |
| fungi | thursday is probably fair game but would be nice to keep things reasonably stable in case of any day-after emergencies | 16:43 |
| clarkb | ya iirc even after the move to wednesday for the bulk of the work we've tried to keep thursday sane for stragglers | 16:44 |
| *** jpena is now known as jpena|off | 16:45 | |
| *** bhavikdbavishi has joined #zuul | 16:47 | |
| openstackgerrit | Merged openstack-infra/nodepool master: Fix race in test_handler_poll_session_expired https://review.openstack.org/650385 | 17:21 |
| *** jamesmcarthur_ has joined #zuul | 17:46 | |
| *** jamesmcarthur has quit IRC | 17:49 | |
| *** electrofelix has quit IRC | 17:54 | |
| openstackgerrit | Clark Boylan proposed openstack-infra/zuul-jobs master: Use /etc/hosts with docker registries to handle ipv6 https://review.openstack.org/650968 | 18:16 |
| clarkb | Shrews: fungi mordred logan- ^ that takes the use /etc/hosts idea and runs with it | 18:16 |
| clarkb | I may be overapplying the /etc/hosts edits (like multiple times in a single role as an example. I haven't looekd closely enough to see if it is safe to do it less often) | 18:17 |
| fungi | yeah, that is a lot of duplication, but can be simplified once proven to work | 18:19 |
| clarkb | the use of regex there should mean it is idempotent if over eager | 18:19 |
| *** Shrews has quit IRC | 18:20 | |
| *** Shrews has joined #zuul | 18:20 | |
| pabelanger | clarkb: yah, I wouldn't want to change that process of opendev update to zuul commit for actually release. But, my brain is thinking, with a beta / rc, that might also provide additional info to our release process too, since opendev can't catch all the use cases. | 18:20 |
| *** gundalow has quit IRC | 18:20 | |
| *** dcastellani has quit IRC | 18:20 | |
| *** PrinzElvis has quit IRC | 18:20 | |
| *** dcastellani has joined #zuul | 18:21 | |
| *** PrinzElvis has joined #zuul | 18:21 | |
| *** gundalow has joined #zuul | 18:21 | |
| pabelanger | I'm trying real hard not to get into the process of backporting patches to some packaging tool , and right now pypi is my source of truth for zuul | 18:21 |
| clarkb | pabelanger: why not use git? | 18:22 |
| pabelanger | and, for the most part, this isn't an issue. We've generally been good a doing zuul releases when asked for one | 18:22 |
| pabelanger | clarkb: I'd need to have build tools on servers to build the wheel, right now I don't have everything installed on the server | 18:23 |
| *** AJaeger has quit IRC | 18:27 | |
| *** sshnaidm is now known as sshnaidm|off | 18:38 | |
| *** AJaeger has joined #zuul | 18:40 | |
| *** jamesmcarthur_ has quit IRC | 19:04 | |
| *** bhavikdbavishi has quit IRC | 19:11 | |
| *** hashar has quit IRC | 19:12 | |
| *** hashar has joined #zuul | 19:13 | |
| fungi | we could start publishing prebuilt wheels too i suppose, if that's desired | 19:22 |
| fungi | (for branch tip) | 19:22 |
| *** jamesmcarthur has joined #zuul | 19:34 | |
| *** jamesmcarthur_ has joined #zuul | 19:35 | |
| openstackgerrit | Clark Boylan proposed openstack-infra/zuul-jobs master: Use /etc/hosts with docker registries to handle ipv6 https://review.openstack.org/650968 | 19:37 |
| clarkb | hopefully that passes linting now | 19:38 |
| *** jamesmcarthur has quit IRC | 19:39 | |
| *** jamesmcarthur_ has quit IRC | 19:45 | |
| Shrews | clarkb: looks pretty good to me. maybe we should have a zuul or nodepool review depend on it and recheck until it hits an ipv6 enabled host? | 19:50 |
| clarkb | Shrews: I don't think that will work because the jobs that use the role are in a trusted repo | 20:02 |
| clarkb | otherwise that would be ++ good | 20:02 |
| Shrews | clarkb: oh right | 20:02 |
| *** jamesmcarthur has joined #zuul | 20:11 | |
| Shrews | clarkb: testing the ipaddr filter you use there... i cannot get that to ever be 'true'. http://paste.openstack.org/show/749024/ | 20:18 |
| Shrews | maybe i'm testing that wrong? | 20:19 |
| clarkb | shrews according to the docs if you give it an ipv4 or ipv6 address you get it back out the other sidr anything else results in false | 20:21 |
| logan- | good catch Shrews | 20:22 |
| clarkb | maybe it is broken or the bool filter isnt what we want to get string to true | 20:22 |
| Shrews | that second line of the paste doesn't show that :( | 20:22 |
| logan- | {{ true if '127.0.0.1' | ipaddr else false }} works | 20:22 |
| clarkb | could compare the input to itself instead of |bool == trye | 20:23 |
| clarkb | addr | ipaddr == addr | 20:23 |
| logan- | that would probably break with v6 depending on address compression | 20:24 |
| logan- | when: "{{ (true if '127.0.0.1' | ipaddr else false) | bool }}" | 20:24 |
| Shrews | we could use the ipv6() alias and just modify hosts for those values | 20:24 |
| logan- | (strike the {{ }} since it is a conditional) | 20:25 |
| *** pcaruana has quit IRC | 20:25 | |
| clarkb | I dont want ipv6 to be different | 20:26 |
| clarkb | will make testing the general case harder | 20:26 |
| Shrews | *nod* | 20:26 |
| clarkb | does when: addr | ipaddr work? | 20:29 |
| mordred | when the moon is full? | 20:29 |
| mordred | clarkb: https://docs.ansible.com/ansible/latest/user_guide/playbooks_filters_ipaddr.html | 20:30 |
| mordred | clarkb: it says when used by itself it's true if it's an ipaddr, false if not | 20:30 |
| clarkb | it says it returns the input value if true | 20:33 |
| mordred | oh.yeah - fair | 20:34 |
| mordred | also - we can use jinja filters in when: conditions? | 20:34 |
| mordred | clarkb: oh - but a value is going to evaluate to true ina when | 20:35 |
| mordred | and it'll be false if it's not one | 20:35 |
| clarkb | mordred: ok that was what I wasn't sure on | 20:35 |
| mordred | so when: foo | ipaddr _should_ work | 20:35 |
| clarkb | so when: addr | ipaddr should be fine then | 20:35 |
| mordred | yeah | 20:35 |
| mordred | other than verifying that we can use jinja on when lines | 20:36 |
| clarkb | and when: not ( addr | ipaddr ) ? | 20:36 |
| clarkb | I'm pretty sure you can it is implicit jinja in when | 20:36 |
| mordred | yeah. one would assume | 20:36 |
| mordred | cool | 20:36 |
| logan- | not ('google.com' | ipaddr) lgtm | 20:36 |
| logan- | and '127.0.0.1' | 20:37 |
| Shrews | ah 'when' is the magic there, then | 20:37 |
| logan- | even without the when, it seems like $ ansible localhost -m debug -a msg="{{ not ('127.0.0.1' | ipaddr) }}" evaluates differently than with the bool filter | 20:38 |
| openstackgerrit | Clark Boylan proposed openstack-infra/zuul-jobs master: Use /etc/hosts with docker registries to handle ipv6 https://review.openstack.org/650968 | 20:41 |
| clarkb | I think ^ should address it now | 20:41 |
| Shrews | yeah, that makes sense to me | 20:47 |
| Shrews | clarkb: ^^ | 20:47 |
| Shrews | thx | 20:48 |
| clarkb | and thank you for reviewing it :) | 20:49 |
| mordred | clarkb: +2 from me too - but I'm not going to +3 it because I'm AFK in a few minutes | 21:01 |
| mordred | +3 at will | 21:01 |
| clarkb | ok that might be a tomorrow item for me as I've buiilt up quite the backlog of things to do already today | 21:01 |
| pabelanger | if people are doing reviews, https://review.openstack.org/650275 is a minor improvement to xterm.js for non-mouse users. Sets focus so our pgup/pgdn keys work, otherwise need to clicked xterm.js first | 21:02 |
| * mordred wishes he had pgup/pgdn keys still | 21:10 | |
| mordred | pabelanger: +2 from me | 21:11 |
| pabelanger | thanks! | 21:12 |
| pabelanger | So, not directly a zuul question, but does anybody know why ansible process would be killed with signal? specifically SIGPIPE? | 21:13 |
| clarkb | pabelanger: with the current code if I click in the xterm window and then pgup/pgdown it doesn't page for me | 21:13 |
| clarkb | now to test your chnage | 21:13 |
| pabelanger | ever so often I get the following failures: https://logs.zuul.ansible.com/periodic-1hr/github.com/ansible-network/windmill-config/master/windmill-config-deploy/8715067/bastion/ara-report/result/49493489-9d84-4637-9375-ad44cba0ad17/ | 21:13 |
| pabelanger | clarkb: shift+pgup? | 21:13 |
| clarkb | oh do I need shift? | 21:13 |
| clarkb | heh I guess it really is an xterm | 21:14 |
| pabelanger | yah, I need that for chrome | 21:14 |
| clarkb | ya its because that is the xterm behavior :) | 21:14 |
| pabelanger | +1 | 21:14 |
| clarkb | https://github.com/docker/distribution/blob/749f6afb4572201e3c37325d0ffedb6f32be8950/reference/regexp.go#L18 is why docker push fails on that ipv6 addr fwiw | 21:35 |
| clarkb | I'm working on filing a bug now so that we don't forget and hopefully in the future we can drop the /etc/hosts hack | 21:35 |
| openstackgerrit | Merged openstack-infra/zuul-jobs master: Add retries to promote-docker-image https://review.openstack.org/644911 | 21:41 |
| SpamapS | tobiash_: wondering if you've done much with the AWS driver. The lack of respect for max-servers means every time we get busy, we get NODE_FAILURE storms. | 22:06 |
| * SpamapS struggling to get back to the update to the driver | 22:07 | |
| SpamapS | Also wondering if there's a way to make those failures more retryable. | 22:07 |
| openstackgerrit | Merged openstack-infra/zuul master: Set focus on xtem.js to terminal https://review.openstack.org/650275 | 22:09 |
| clarkb | SpamapS: by default they should be retried 3 times per provider | 22:10 |
| clarkb | though maybe that is provider specific? I know that is what the openstack driver does | 22:10 |
| SpamapS | clarkb: yes, they are retried 3 times, fairly rapidly (1s pause) and then fail. | 22:14 |
| SpamapS | I'd prefer that they just give back the request. | 22:14 |
| SpamapS | But guessing there's some problem with that. | 22:14 |
| SpamapS | The problem I have is that there's a per-region max-instances, and when we hit it.. #FAIL | 22:14 |
| clarkb | why not implement max-servers in the aws driver? | 22:15 |
| clarkb | seems like that would be the best thing | 22:15 |
| SpamapS | that's what I have to do | 22:15 |
| SpamapS | with all the time I don't have, it should be simple. ;) | 22:15 |
| SpamapS | Also it will have other problems I'd like to figure out. I have a provider-per-AZ, but there's a regional max. Sometimes AZ's run out of instance types, so I'd prefer a single provider, with a single max-servers, and some kind of retry-on-another-az logic. | 22:16 |
| *** tobiash_ has quit IRC | 22:17 | |
| clarkb | https://github.com/moby/moby/issues/39033 is the docker bug | 22:19 |
| clarkb | there is a chance that a minor update to the regex would fix this, but I don't really feel like being responsible for sorting that out. I guess I could write the regex update and ask them to shepherd from there | 22:19 |
| *** tobiash has joined #zuul | 22:21 | |
| clarkb | mordred: Shrews logan- on my change one thing we should keep in mind is it may affect the ipv4 jobs | 22:28 |
| clarkb | which is why i didn't just merge it considering "its already broken" | 22:28 |
| clarkb | but I'll try to approve that tomorrow morning and watch it from there | 22:28 |
| *** persia has quit IRC | 22:30 | |
| *** persia has joined #zuul | 22:31 | |
| *** hashar has quit IRC | 23:02 | |
| *** mgoddard has quit IRC | 23:14 | |
| *** mgoddard has joined #zuul | 23:15 | |
| *** jamesmcarthur has quit IRC | 23:22 | |
| *** rlandy has quit IRC | 23:52 | |
| *** jamesmcarthur has joined #zuul | 23:56 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!