| *** armstrongs has joined #zuul | 00:04 | |
| *** mattw4 has quit IRC | 00:09 | |
| *** tosky has quit IRC | 00:09 | |
| *** armstrongs has quit IRC | 00:14 | |
| *** igordc has quit IRC | 00:58 | |
| *** dcastellani has quit IRC | 01:12 | |
| *** dcastellani has joined #zuul | 01:12 | |
| *** portdirect has quit IRC | 01:12 | |
| *** kmalloc has quit IRC | 01:13 | |
| *** kmalloc has joined #zuul | 01:13 | |
| *** mordred has quit IRC | 01:14 | |
| *** webknjaz has quit IRC | 01:14 | |
| *** mordred has joined #zuul | 01:15 | |
| *** portdirect has joined #zuul | 01:16 | |
| *** webknjaz has joined #zuul | 01:16 | |
| *** rfolco|doctor is now known as rfolco | 01:18 | |
| *** rfolco has quit IRC | 01:34 | |
| *** rlandy has quit IRC | 02:04 | |
| *** bhavikdbavishi has joined #zuul | 02:25 | |
| *** swest has quit IRC | 02:35 | |
| *** swest has joined #zuul | 02:49 | |
| *** raukadah is now known as chandankumar | 04:11 | |
| *** swest has quit IRC | 05:23 | |
| *** evrardjp has quit IRC | 05:34 | |
| *** bhavikdbavishi has quit IRC | 05:34 | |
| *** evrardjp has joined #zuul | 05:34 | |
| *** swest has joined #zuul | 06:04 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Evaluate CODEOWNERS settings during canMerge check https://review.opendev.org/644557 | 07:52 |
|---|---|---|
| *** Defolos has joined #zuul | 07:58 | |
| *** armstrongs has joined #zuul | 07:58 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add --validate-tenants option to zuul scheduler https://review.opendev.org/542160 | 08:00 |
| *** armstrongs has quit IRC | 08:08 | |
| *** bhavikdbavishi has joined #zuul | 08:09 | |
| *** tosky has joined #zuul | 08:27 | |
| *** avass has quit IRC | 08:30 | |
| *** avass has joined #zuul | 08:30 | |
| *** bhavikdbavishi has quit IRC | 08:36 | |
| *** jpena|off is now known as jpena | 08:43 | |
| *** avass has quit IRC | 08:50 | |
| *** dpawlik has joined #zuul | 09:17 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Authorization rules: add templating https://review.opendev.org/705193 | 09:37 |
| openstackgerrit | Jan Kubovy proposed zuul/zuul master: Make test setup_config more pure https://review.opendev.org/709084 | 10:05 |
| openstackgerrit | Jan Kubovy proposed zuul/zuul master: Make test setup_config more pure https://review.opendev.org/709084 | 10:06 |
| openstackgerrit | Jan Kubovy proposed zuul/zuul master: Make test setup_config more pure https://review.opendev.org/709084 | 10:15 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Add --validate-tenants option to zuul scheduler https://review.opendev.org/542160 | 10:28 |
| *** avass has joined #zuul | 11:22 | |
| *** bhavikdbavishi has joined #zuul | 12:36 | |
| *** jpena is now known as jpena|lunch | 12:42 | |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Add ensure-bazelisk role https://review.opendev.org/708899 | 12:51 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Add ensure-java role https://review.opendev.org/708901 | 12:52 |
| *** rfolco has joined #zuul | 12:53 | |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Add ensure-bazelisk role https://review.opendev.org/708899 | 12:56 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Add ensure-java role https://review.opendev.org/708901 | 12:56 |
| *** rlandy has joined #zuul | 12:59 | |
| *** bhavikdbavishi has quit IRC | 13:02 | |
| tobiash | corvus: do you have any objection about removing the repo state from the 'Merge completed for item' log line? We have repos that produce >25kb for just this single log line | 13:03 |
| *** Goneri has joined #zuul | 13:09 | |
| mordred | tobiash: you have amazing repos | 13:10 |
| tobiash | mordred: we already filter out most unprotected branches but there are a ton of tags left ;) | 13:11 |
| openstackgerrit | Felix Edel proposed zuul/zuul master: Dequeue items via buildset uuid https://review.opendev.org/709135 | 13:17 |
| mordred | tobiash: :) | 13:18 |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: WIP: test install-docker all-platforms https://review.opendev.org/709139 | 13:37 |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: WIP: test install-docker all-platforms https://review.opendev.org/709139 | 13:39 |
| *** dpawlik has quit IRC | 13:40 | |
| *** jpena|lunch is now known as jpena | 13:42 | |
| *** dpawlik has joined #zuul | 13:48 | |
| *** rishabhhpe has joined #zuul | 13:54 | |
| EmilienM | hello Zuul :) | 13:57 |
| *** Goneri has quit IRC | 13:57 | |
| EmilienM | today I tried to put 2 gerrit IDs in the search bar on https://zuul.openstack.org/status | 13:58 |
| EmilienM | 2 IDs work fine, the filter works fine; however with 3 it doesn't work anymore, it only shows the first in the list | 13:58 |
| *** dpawlik has quit IRC | 14:00 | |
| mnaser | EmilienM: strange, i just tried "709089,709094,709101" and it worked fine | 14:01 |
| EmilienM | ah let me try with comams | 14:01 |
| EmilienM | commas* | 14:02 |
| EmilienM | mnaser: ah it works fine, thanks! | 14:03 |
| mnaser | should probably have that somehow clarified | 14:03 |
| EmilienM | mnaser: interesting, if you remove the commas, it works for 2 IDs | 14:03 |
| EmilienM | but 3, it only show the first ID | 14:03 |
| mnaser | EmilienM: oh yeah, i see what you're seeing | 14:04 |
| mnaser | `let filters = filter.replace(/ +/, ',').split(',')` | 14:07 |
| mnaser | i bet that regex replaces only once because g is probably missing.. | 14:08 |
| mnaser | yep thats it | 14:08 |
| openstackgerrit | Mohammed Naser proposed zuul/zuul master: ui: convert all spaces to commas for filter https://review.opendev.org/709147 | 14:10 |
| mnaser | EmilienM: that should be the fix ^ | 14:10 |
| EmilienM | ah wow :) | 14:10 |
| Shrews | mnaser: it seems that line also appears in Jobs.jsx | 14:11 |
| mnaser | Shrews: ah yes, good catch | 14:12 |
| openstackgerrit | Mohammed Naser proposed zuul/zuul master: ui: convert all spaces to commas for filter https://review.opendev.org/709147 | 14:12 |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Don't fetch pull request twice for status event https://review.opendev.org/709149 | 14:14 |
| openstackgerrit | Felix Edel proposed zuul/zuul master: WIP: Dequeue items via buildset uuid https://review.opendev.org/709135 | 14:24 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul master: prometheus: add options to start the server and process collector https://review.opendev.org/599209 | 14:39 |
| *** rfolco has quit IRC | 14:47 | |
| *** rfolco has joined #zuul | 14:54 | |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: WIP: test install-docker all-platforms https://review.opendev.org/709139 | 14:56 |
| mordred | corvus: you may have noticed, but I updated your two zuul-jobs changes with the java stuff | 14:57 |
| corvus | mordred: yes, thanks | 14:57 |
| mordred | (that was mostly a "if you need to update them, be sure to download again first" :) ) | 14:57 |
| openstackgerrit | Merged zuul/zuul master: ui: convert all spaces to commas for filter https://review.opendev.org/709147 | 15:13 |
| AJaeger | mordred, corvus, +2A - so, if you want to update, better WIP now or make a followup change ;) | 15:16 |
| *** bhavikdbavishi has joined #zuul | 15:19 | |
| corvus | nah they lgtm | 15:20 |
| openstackgerrit | David Shrewsbury proposed zuul/zuul master: Add nodesets API route https://review.opendev.org/605877 | 15:23 |
| Shrews | mordred: fixed ^^ for you | 15:24 |
| corvus | tristanC, Shrews, mordred, tobiash: see my reply to tristanC's comment on https://review.opendev.org/621479 | 15:28 |
| Shrews | corvus: i think that's one of the things we don't know until we know it :/ | 15:28 |
| Shrews | my comment was more of a "let's watch and learn" approach | 15:29 |
| corvus | (the few benchmarks i have been able to find suggest that with the client numbers i would expect (<100), performance would be similar except that zk will use 4x ram -- so if we store 1GB of data, we'll need 4GB of ram; mostly from this page: https://coreos.com/blog/performance-of-etcd.html) | 15:31 |
| openstackgerrit | Merged zuul/zuul-jobs master: Add ensure-bazelisk role https://review.opendev.org/708899 | 15:31 |
| corvus | based on that, i'm not spooked -- i think i agree with Shrews that we can take a wait-and-see approach, and if we run into problems, revisit it. but if someone see's something that does spook them, maybe we should do our own benchmarks with simulated data first... | 15:32 |
| *** Goneri has joined #zuul | 15:36 | |
| mnaser | i'd be in favour of running etcd personally | 15:37 |
| mnaser | it's a lot easier to get started with, operationally it's also simpler imho | 15:38 |
| corvus | mnaser: i don't think zk has been a problem operationally. | 15:38 |
| mnaser | eh it's a lot less trivial to run imho than etcd, but also there's the fact that etcd will probably continue to see improvements because of its heavy usage insidek 8s | 15:39 |
| mnaser | k8s* | 15:39 |
| corvus | mnaser: for zuul it has generally not been an issue | 15:39 |
| *** jpena is now known as jpena|off | 15:39 | |
| mnaser | forgive me to sound like a bit of a broken record but i'm not a huge fan of running java things | 15:40 |
| *** dpawlik has joined #zuul | 15:41 | |
| *** jpena|off is now known as jpena | 15:42 | |
| corvus | we've had the "etcd vs zookeeper" discussion before, and we have come to a conclusion, so "it's in java" and "k8s uses it" are all things that were true the last time we had it. if we're going to reopen the discussion, i'd like it to be based on something that's different. ie, "this proposed usage pattern will suffer under zk and shine under etcd" or "the zk has folded and does not maintain their software" | 15:42 |
| *** rishabhhpe has quit IRC | 15:43 | |
| tristanC | corvus: it seems like zookeeper is no longer popular and it looks like in maintainance mode. however, it seems to work perfectly fine and switching to another system like etcd is going to cause a lot of other problems we should consider in the evaluation | 15:43 |
| *** plaurin has joined #zuul | 15:45 | |
| corvus | tristanC: yes, that's how i feel. if we were starting from scratch, i think we would chose etcd at this point (i believe it now meets our technical requirements). but changing will set back having an HA scheduler by many months (whether we do that before or after we start the work). | 15:46 |
| tristanC | fwiw, here is a kafka dev thread about zookeeper removal: https://lists.apache.org/thread.html/cce5313ebe72bde34bf0da3af5a1723db3ee871667b1fd8edf2ee7ab@%3Cdev.kafka.apache.org%3E | 15:48 |
| openstackgerrit | Merged zuul/zuul-jobs master: Add ensure-java role https://review.opendev.org/708901 | 15:49 |
| corvus | tristanC: interestingly, that is not replacing it with any existing software, they are implementing their own raft-consensus system to get unique features. | 15:50 |
| *** avass has quit IRC | 15:50 | |
| mnaser | looks like going based on the current nodepool code, it would likely need to be driver-ified then have a zk/etcd driver for transition | 15:52 |
| mnaser | or rather there is nodepool.zk | 15:53 |
| tristanC | mnaser: the cost of such abstraction can be very high as nodepool currently uses some unique feature of zk, e.g. unique sequence node and the zk lock api | 15:54 |
| mnaser | ah, please forgive my *very* ignorant comments then, as i don't understand the details of all of this | 15:54 |
| fungi | yeah, i don't question our choice to go with zk over etcd. at the time we needed somewhere to do this zk had the features we needed and etcd did not. etcd seems to have caught up, but waiting for it would have set our own development back considerably. now the bar for switching is that the pain of a transition needs to be outweighed by the benefits to our community | 15:55 |
| tristanC | iirc, the locking api is most problematic as it's different in etcd. iiuc in zk you have one lock per object that get's auto attached to the session, and in etcd you have one lock per session where you attach objects | 15:56 |
| clarkb | there was a recent thing about how poorly etcd locks are implemented. I'm not sureif zk suffers the same issue (could be zab and raft suffer thr same issue when reaching consensus I'm not sure). I think it does point out that we may have to be careful about behavior and assumptions though | 15:56 |
| fungi | when we can demonstrate we've reached that tipping point, i'm fine with the idea of trying to transition | 15:56 |
| clarkb | tristanC: and in etcd multiple sessions can hold an exclhsive lock | 15:57 |
| tristanC | clarkb: is this the thing you are talking about? https://jepsen.io/analyses/etcd-3.4.3 | 15:57 |
| clarkb | tristanC: ya thats the one | 15:58 |
| pabelanger | I would agree, running zookeeper for zuul has almost been zero cost; it just works. I'd not be a fan of replacing it over something else at this point. | 15:58 |
| tristanC | pabelanger: i wouldn't say zero cost. didn't we found that for nodepool, you need 1GB of zk ram for every 10 nodes? | 15:59 |
| fungi | and it's packaged by distros, so i haven't actually had to think about "installing a java application" much less touch java source code simply by relying on zk | 16:00 |
| mnaser | its neat that the CNCF is funding research like that | 16:00 |
| fungi | agreed, i bet that wasn't cheap | 16:01 |
| pabelanger | tristanC: sorry, I ment operational cost. You still need 3 properly sided servers for cluster | 16:01 |
| pabelanger | sized* | 16:01 |
| tristanC | i think it's a valid concern that adding all the scheduler data in zk can become an issue, but i also think that at this point, we might as wait and see, and refactor the zk.py module at once we know the exact limitation | 16:02 |
| tristanC | we might as well* wait and see, and refactor the zk.py module once* we know the exact limitations* | 16:03 |
| clarkb | fungi: they also publish a tarball with init script in it that just works. I use that brcause tumbleweed doesnt paclage zk | 16:03 |
| fungi | clarkb: ooh, neat | 16:03 |
| fungi | so just splat out the tarball, edit config, run the initscript, and you have zk? | 16:04 |
| clarkb | yes | 16:04 |
| tristanC | clarkb: does the script can setup authentication too? | 16:04 |
| clarkb | tristanC: you would need to edit the config yourself for that | 16:05 |
| clarkb | but the software is there and can be started and stopped | 16:05 |
| fungi | presumably you'd need to also generate your credentials for that | 16:06 |
| tristanC | clarkb: i find that adding acl settings to ZK to be difficult and britle, e.g. it needs one jaas file, an option in zk.conf, and a jvm flag . Then the only way to verify it works is to edit a zk node with a wrong acl... | 16:06 |
| tristanC | compared to etcd where you can just give it tls cert using the cli, similarly to gearman, and you have an authenticated only service | 16:08 |
| fungi | yeah, injecting files into a jvm is annoying, i didn't realize zk required that (or is that only if you want to use self-signed certs?) | 16:09 |
| fungi | (or a custom ca?) | 16:09 |
| tristanC | fwiw i don't think setup cost or jvm runtime is relevant for the scaled-out discussion, that's a one time cost we can fix using doc. | 16:10 |
| fungi | if the problem is updating the jvm's baked-in trust chain, maybe letsencrypt is an alternative solution | 16:10 |
| *** rfolco is now known as rfolco|eats | 16:11 | |
| tristanC | fungi: here is the zk auth setup procedure: https://review.opendev.org/#/c/619155/23/tools/test-setup.sh | 16:11 |
| tristanC | fungi: and the jaas file: https://review.opendev.org/#/c/619155/23/nodepool/tests/fixtures/zookeeper/auth.conf | 16:12 |
| fungi | okay, so that's not altering files within the jvm. doesn't look terrible | 16:13 |
| tristanC | fungi: that's not too bad, it's just a terrible operator experience because the service never fail to start and you can't tell if auth is activated until you actually try to write a node with bad acl | 16:14 |
| fungi | yep, that does seem suboptimal | 16:14 |
| fungi | doesn't seem on its own to be bad enough to warrant replacing zk though, if that's the only substantive complaint we have | 16:15 |
| fungi | but if folks are seriously discussing the idea, then we likely need a list of what the problems we have with zk are and which of them could be solved by etcd | 16:16 |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Enable testing of install-docker on multiple platforms https://review.opendev.org/709139 | 16:16 |
| openstackgerrit | Merged zuul/zuul master: Disable misfire grace time of apscheduler job https://review.opendev.org/707842 | 16:18 |
| tristanC | fungi: yeah no, i wouldn't consider auth configuration a good reason to remove zk. it's just a consequence of zk not being designed from the ground up with auth | 16:18 |
| clarkb | fwiw sasl will be in the next release | 16:19 |
| clarkb | that should address tristanC's concern | 16:19 |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Enable testing of install-docker on multiple platforms https://review.opendev.org/709139 | 16:20 |
| tristanC | clarkb: iiuc tobiash, we would still need to use https://review.opendev.org/#/c/619155/23/nodepool/cmd/update_zk_auth.py | 16:21 |
| clarkb | tristanC: it depends on whether or not you can configure zk to require auth when using sasl | 16:22 |
| clarkb | I'm not sure if you can or not | 16:22 |
| clarkb | but it will simplify the authentication setup greatly as you'd just set the certs aiui | 16:22 |
| clarkb | since sasl occurs at connection start time you shouldn't need to configure acls if you have single purpose clusters and require sasl (if that is possible) | 16:23 |
| tristanC | clarkb: tobiash mentioned that setting: https://cwiki.apache.org/confluence/display/ZOOKEEPER/ZooKeeper+SSL+User+Guide | 16:23 |
| openstackgerrit | Merged zuul/zuul master: Spec for allowing circular dependencies https://review.opendev.org/643309 | 16:24 |
| clarkb | tristanC: requireClientAuthScheme=sasl | 16:30 |
| clarkb | that is the config option to set then it should work just like gearman aiui | 16:30 |
| clarkb | if you don't multitenant the cluster you don't need to touch acls at all | 16:30 |
| clarkb | the docs I found say that applies to all connections so I think it affects server to server and client to server | 16:32 |
| fungi | headed out for food, back as soon as i can be | 16:32 |
| fungi | er, wrong channel, sorry | 16:32 |
| clarkb | (you do have to enable the use of acls though) | 16:32 |
| *** rfolco|eats is now known as rfolco | 16:33 | |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: WIP: Fix install-docker on centos-8 https://review.opendev.org/709184 | 16:36 |
| *** zenkuro has joined #zuul | 16:37 | |
| *** bhavikdbavishi has quit IRC | 16:41 | |
| zenkuro | hello folks, I was doing zuul from scratch by https://zuul-ci.org/docs/zuul/howtos/zuul-from-scratch.html and I am not able to see builds and buildsets, and have issues with configuring gerrit + Im missing the idea/behind adding mysql(and how to do it properly) | 16:41 |
| clarkb | zenkuro: you'll need a database (mysql) in order to have the builds and buildset data in the dashboard. | 16:42 |
| clarkb | zenkuro: I think that means we should maybe start with the db config? | 16:43 |
| clarkb | zenkuro: can you share what you've got for the zuul config so far? then we can work from what you've already got configured for mysql | 16:44 |
| zenkuro | Oh... I see. If someone can supervise me on digging how to make zuul from scratch equal to zuul quick start. I can improve documentation | 16:44 |
| clarkb | (feel free to remove secret data as necessary) | 16:44 |
| corvus | zenkuro: it looks like zuul-from-scratch doesn't cover the database. the quick-start does include a working database: https://zuul-ci.org/docs/zuul/tutorials/quick-start.html#quick-start but doesn't describe how to set it up. here's the zuul config file: https://opendev.org/zuul/zuul/src/branch/master/doc/source/examples/etc_zuul/zuul.conf | 16:44 |
| clarkb | ah | 16:44 |
| corvus | zenkuro: that would be great; i think zuul-from-scratch just needs database installation instructions, creating an initial database user, then an update to the zuul.conf file to add the 'connection "mysql"' part | 16:45 |
| corvus | zenkuro: the final piece is making sure that the pipeline has a "mysql" reporter in it | 16:45 |
| corvus | zenkuro: (lik this file in the quick-start: https://opendev.org/zuul/zuul/src/branch/master/doc/source/examples/zuul-config/zuul.d/pipelines.yaml ) | 16:46 |
| openstackgerrit | James E. Blair proposed zuul/zuul master: Remove implemented specs https://review.opendev.org/701435 | 16:49 |
| zenkuro | clarkb: zuul.conf https://softwarefactory-project.io/paste/show/1741/ | 16:54 |
| *** mattw4 has joined #zuul | 16:56 | |
| clarkb | zenkuro: using the quickstart example config that corvus linked you'll need to add a section like https://opendev.org/zuul/zuul/src/branch/master/doc/source/examples/etc_zuul/zuul.conf#L28-L31 | 16:56 |
| zenkuro | corvus: so all I need is to install mysql and add zuul user?(config as it is configured in quick start)? | 16:56 |
| clarkb | zenkuro: you might have to create the database too. But yes zuul should handle the rest once it connects | 16:57 |
| zenkuro | clarkb: thanks, I will try it out. | 16:59 |
| *** Defolos has quit IRC | 17:07 | |
| *** tosky has quit IRC | 17:23 | |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Fix install-docker on centos-8 https://review.opendev.org/709184 | 17:30 |
| openstackgerrit | Clark Boylan proposed zuul/zuul-website master: Add redirect for /docs/zuul/user/ https://review.opendev.org/709195 | 17:32 |
| clarkb | I noticed ^ when fiddling with some log parsing tools | 17:32 |
| clarkb | another common one is _static/logo.svg but I think that may just be indexers assuming that is where a logo lives? | 17:33 |
| clarkb | we could add a redirect for it too if we want to point them in the right direction | 17:33 |
| clarkb | oh and docs/_static/logo.svg | 17:33 |
| *** evrardjp has quit IRC | 17:34 | |
| *** evrardjp has joined #zuul | 17:35 | |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Enable testing of install-docker on multiple platforms https://review.opendev.org/709139 | 17:39 |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Fix install-docker on centos-8 https://review.opendev.org/709184 | 17:40 |
| *** phildawson has quit IRC | 17:42 | |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Fix install-docker on centos-8 https://review.opendev.org/709184 | 17:43 |
| *** jpena is now known as jpena|off | 17:47 | |
| *** plaurin has quit IRC | 17:58 | |
| *** jamesmcarthur has joined #zuul | 17:59 | |
| openstackgerrit | Merged zuul/zuul master: Remove implemented specs https://review.opendev.org/701435 | 18:02 |
| corvus | tristanC: i've run into a weird problem: i installed k8s with kind and hooked it up to a nodepool launcher. i'm running a test job, and ansible is stuck at the gathering facts stage. according to ps, it's running this command: | 18:10 |
| corvus | /usr/local/bin/kubectl exec -i pod-fedora -- /bin/sh -c /bin/sh -c 'echo PLATFORM; uname; echo FOUND; command -v '"'"'/usr/bin/python'"'"'; command -v '"'"'python3.7'"'"'; command -v '"'"'python3.6'"'"'; command -v '"'"'python3.5'"'"'; command -v '"'"'python2.7'"'"'; command -v '"'"'python2.6'"'"'; command -v '"'"'/usr/libexec/platform-python'"'"'; command -v '"'"'/usr/bin/python3'"'"'; command -v | 18:10 |
| corvus | '"'"'python'"'"'; echo ENDFOUND && sleep 0' | 18:10 |
| corvus | tristanC: it *looks* like it has an extra "/bin/sh -c", and if i remove that and run it manually, it works | 18:11 |
| corvus | have you seen anything like this before? | 18:11 |
| *** masterpe has quit IRC | 18:12 | |
| tristanC | corvus: i have seen the dual /bin/sh -c before and it didn't caused issue. | 18:13 |
| tristanC | though i don't know how kind work | 18:14 |
| *** jamesmcarthur has quit IRC | 18:14 | |
| corvus | maybe that's a red herring then | 18:14 |
| tristanC | i've seen that when implementing https://pagure.io/software-factory/k1s , and gather_fact worked | 18:16 |
| corvus | hrm, i just ran the same command against a pod in a "real" k8s cluster and got the same behavior (hangs with 2 bin/sh, works with 1) | 18:16 |
| *** plaurin has joined #zuul | 18:17 | |
| tristanC | what's the container image? | 18:18 |
| corvus | tristanC: i've tried with fedora:28 and ubuntu:bionic | 18:20 |
| *** masterpe has joined #zuul | 18:21 | |
| *** jamesmcarthur has joined #zuul | 18:22 | |
| *** jamesmcarthur_ has joined #zuul | 18:23 | |
| *** jamesmcarthur has quit IRC | 18:26 | |
| *** masterpe has quit IRC | 18:30 | |
| *** openstackstatus has quit IRC | 18:33 | |
| *** armstrongs has joined #zuul | 18:37 | |
| *** armstrongs has quit IRC | 18:47 | |
| *** jamesmcarthur_ has quit IRC | 18:52 | |
| openstackgerrit | Andreas Jaeger proposed zuul/zuul-jobs master: Enable testing of install-docker on multiple platforms https://review.opendev.org/709139 | 18:55 |
| *** jamesmcarthur has joined #zuul | 18:59 | |
| *** jamesmcarthur_ has joined #zuul | 19:01 | |
| *** masterpe has joined #zuul | 19:02 | |
| *** jamesmcarthur has quit IRC | 19:04 | |
| corvus | tristanC: even stranger, i can not replicate the problem if i manually run 'ansible-playbook' with the job's inventory and kube config | 19:07 |
| corvus | tristanC: aha! i think it's something to do with my kubectl which comes from snap | 19:07 |
| corvus | i think maybe i will stop using snap altogether. | 19:08 |
| *** igordc has joined #zuul | 19:10 | |
| corvus | yep. all better now. sorry. that was ridiculous. | 19:11 |
| *** Defolos has joined #zuul | 19:11 | |
| tristanC | hmm, that's odd. glad you figured that out :) | 19:12 |
| Shrews | corvus: was the snap version older or something? | 19:12 |
| fungi | switched from snaps to distro package? | 19:12 |
| corvus | no, snap is some weird thing where it symlinks 'kubectl' to 'snap' and it switches on argv[0] and does "snappy" things before running, and those "snappy" things don't work in bwrap (which is good! because i'm pretty sure they involve updating stuff on the system) | 19:13 |
| Shrews | ah, a snap bwrap flap | 19:14 |
| corvus | so i solved it by doing what everyone else which is download the binary from github. because there are zero other options. | 19:14 |
| corvus | i'm going to go release energy. | 19:15 |
| *** ianychoi_ has joined #zuul | 19:15 | |
| fungi | in the form of a gamma ray burst? | 19:17 |
| * fungi puts on his lead-lined sunglasses | 19:17 | |
| * mordred misses the fun issues | 19:17 | |
| mordred | corvus: the other day I was thinking that maybe we should stop installing kubectl via snap in system-config - but not for a real reason, just because we use the oc download in jobs and we don't really do anything with snap anywhere else | 19:18 |
| *** ianychoi has quit IRC | 19:18 | |
| tristanC | corvus: on centos, you can yum install origin-clients which provides a working kubectl | 19:19 |
| openstackgerrit | Merged zuul/zuul-jobs master: Enable testing of install-docker on multiple platforms https://review.opendev.org/709139 | 19:25 |
| mnaser | corvus: what os do you use? | 19:25 |
| mnaser | i've historically just found it eay to add the upstream repos to my laptop and install kubectl only | 19:25 |
| mordred | mnaser: he "tried with fedora:28 and ubuntu:bionic" | 19:26 |
| mnaser | oh woops | 19:26 |
| mnaser | i should read | 19:26 |
| mordred | mnaser: are there upstream repos that can be used here? | 19:26 |
| mnaser | https://kubernetes.io/docs/tasks/tools/install-kubectl/#install-using-native-package-management | 19:26 |
| *** chandankumar is now known as raukadah | 19:27 | |
| mordred | mnaser: oh cool | 19:27 |
| mnaser | i run debian sid and i use those just fine, i've also ran them on ubuntu and centos with no problems, never tested with fedora, cant imagine they can be too much of an issue | 19:27 |
| mnaser | it is pretty hidden thugh. | 19:27 |
| *** zenkuro has quit IRC | 19:34 | |
| fungi | kubectl is in https://packages.debian.org/sid/amd64/kubernetes-client/filelist but that's only on sid so far | 19:36 |
| fungi | not picked up by ubuntu yet (and likely won't be until it at least migrates to testing) | 19:37 |
| *** jamesmcarthur_ has quit IRC | 19:38 | |
| fungi | looks like it's waiting for these to also enter testing: golang-github-rackspace-gophercloud, golang-golang-x-exp, kubernetes-addon-heapster, skydns | 19:38 |
| mnaser | fungi: that's an ancient verison (1.7) -- i tried working on repackaging it but the other problem is that k8s doesnt build with golang 1.13 yet | 19:38 |
| mnaser | and sid has golang 1.13 so.. | 19:38 |
| fungi | yeah, so much ugh around distro packaging golang stuff | 19:39 |
| mordred | yah | 19:39 |
| mnaser | btw i'd like to ask on some thought around https://review.opendev.org/#/c/708731/8 | 19:39 |
| mnaser | i'm unsure as to how the best way to go about resolving this, or if that test is needed at all anymore (if i have one that doesn't let you do that behaviour..) | 19:40 |
| *** zenkuro has joined #zuul | 19:46 | |
| *** jamesmcarthur has joined #zuul | 19:46 | |
| *** jamesmcarthur has quit IRC | 19:52 | |
| corvus | mnaser: replied | 19:57 |
| *** plaurin has joined #zuul | 20:29 | |
| *** plaurin has quit IRC | 20:43 | |
| *** sgw has quit IRC | 20:46 | |
| *** sgw has joined #zuul | 21:06 | |
| *** rfolco has quit IRC | 21:52 | |
| *** rlandy has quit IRC | 22:10 | |
| *** dpawlik has quit IRC | 22:22 | |
| *** Defolos has quit IRC | 22:28 | |
| *** mattw4 has quit IRC | 22:35 | |
| *** mattw4 has joined #zuul | 22:36 | |
| *** rfolco has joined #zuul | 23:03 | |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Add pause-buildset-registry role https://review.opendev.org/709256 | 23:26 |
| openstackgerrit | Monty Taylor proposed zuul/zuul-jobs master: Add pause-buildset-registry role https://review.opendev.org/709256 | 23:28 |
| *** mattw4 has quit IRC | 23:49 | |
| openstackgerrit | James E. Blair proposed zuul/nodepool master: WIP: k8s: allow port forwarding https://review.opendev.org/709259 | 23:49 |
| openstackgerrit | James E. Blair proposed zuul/zuul master: WIP: Stream output from kubectl pods https://review.opendev.org/709261 | 23:54 |
| corvus | tobiash, tristanC, clarkb, fungi, Shrews, mordred: ^ that should implement the thing that plaurin was asking for that we talked about the other day. i tested it locally. there's a big question in the nodepool change that needs answering. | 23:55 |
| fungi | oh, neat! | 23:55 |
| corvus | (but either way, this does look feasible) | 23:55 |
| fungi | corvus: arbitrary ports on the executor or on the pod? i'm probably just not familiar enough with the kubernetes provider architecture to understand why it would allow it for the zuul executor | 23:59 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!