| openstackgerrit | Clark Boylan proposed openstack-infra/zuul-jobs master: Use mirrors if available when installing OVS on centos https://review.openstack.org/624525 | 00:27 |
|---|---|---|
| *** rlandy has quit IRC | 01:05 | |
| openstackgerrit | Clark Boylan proposed openstack-infra/zuul-jobs master: Use mirrors if available when installing OVS on centos https://review.openstack.org/624525 | 02:11 |
| *** bhavikdbavishi has joined #zuul | 02:48 | |
| openstackgerrit | Clark Boylan proposed openstack-infra/zuul-jobs master: Use mirrors if available when installing OVS on centos https://review.openstack.org/624525 | 04:17 |
| openstackgerrit | Ian Wienand proposed openstack-infra/zuul-jobs master: mirror-workspace-git-repos: Explicitly show HEAD of checked out branches https://review.openstack.org/621840 | 05:06 |
| openstackgerrit | Ian Wienand proposed openstack-infra/zuul-jobs master: Update test-mirror-workspace-git-repos, add test https://review.openstack.org/624575 | 05:06 |
| openstackgerrit | Ian Wienand proposed openstack-infra/zuul-jobs master: Add a note on testing https://review.openstack.org/624578 | 05:44 |
| *** bhavikdbavishi1 has joined #zuul | 06:36 | |
| *** bhavikdbavishi has quit IRC | 06:37 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 06:37 | |
| *** quiquell|off is now known as quiquell | 07:13 | |
| *** openstackgerrit has quit IRC | 07:29 | |
| *** themroc has joined #zuul | 07:57 | |
| *** hashar has joined #zuul | 08:01 | |
| *** bhavikdbavishi has quit IRC | 08:35 | |
| *** openstackgerrit has joined #zuul | 08:51 | |
| openstackgerrit | Jean-Philippe Evrard proposed openstack-infra/zuul-jobs master: Add docker insecure registries feature https://review.openstack.org/624484 | 08:51 |
| *** jpena|off is now known as jpena | 08:55 | |
| *** bhavikdbavishi has joined #zuul | 09:19 | |
| *** bhavikdbavishi has quit IRC | 09:23 | |
| *** bhavikdbavishi has joined #zuul | 09:48 | |
| *** electrofelix has joined #zuul | 10:04 | |
| *** sshnaidm|afk is now known as sshnaidm | 10:05 | |
| *** quiquell is now known as quiquell|brb | 10:19 | |
| *** gtema has joined #zuul | 10:30 | |
| *** quiquell|brb is now known as quiquell | 10:40 | |
| *** gtema has quit IRC | 10:50 | |
| *** bhavikdbavishi has quit IRC | 11:19 | |
| *** rfolco has joined #zuul | 11:30 | |
| *** bhavikdbavishi has joined #zuul | 12:39 | |
| *** jpena is now known as jpena|lunch | 12:43 | |
| evrardjp | a stupid question for this channel: I am doing check/gate on docker image building, and I want to publish the built image into dockerhub. | 13:22 |
| evrardjp | I thought of adding a post job with the dockerhub user/pass secret, which runs a playbook accessing that secret. That playbook will basically run shell commands to docker login/docker push using said secrets | 13:23 |
| pabelanger | evrardjp: yup, that works. There are a few projects in openstack doing that, also here is #zuul we publish images that way too | 13:24 |
| evrardjp | Sadly, I suppose it won't work, as the "post" pipeline is a different pipeline, right, and therefore the target node of the job will not have the image built anymore, right? | 13:24 |
| evrardjp | pabelanger: how can I ensure that said job will run on the same machine as the gate? | 13:25 |
| evrardjp | pabelanger: Or maybe I should do a different way, using a post-run playbook that reaches the secret and runs only on post-merge? | 13:26 |
| pabelanger | evrardjp: can't today, nodepool will give you a new VM to start builds again. | 13:26 |
| pabelanger | let me get job as example | 13:26 |
| evrardjp | pabelanger: so it means that I have to effectively build 3 times: check > gate > post | 13:26 |
| pabelanger | evrardjp: yes | 13:27 |
| pabelanger | there is a way to do artifact handling to build in check or gate, then upload to some static host, then in post download and publish | 13:28 |
| pabelanger | however, really you want to build the artifact post gate, because there is a small chance something could happen that prevented the code from merging, and if then used that artifact in post you no longer using the code that has been merged | 13:28 |
| evrardjp | probably the wrong channel to ask that, but maybe infra has a registry I can use for temp posting? | 13:29 |
| pabelanger | evrardjp: no registry, mostly tarballs.o.o | 13:29 |
| evrardjp | pabelanger: I see what you mean | 13:29 |
| pabelanger | there is work to do a new artifacts pipeline form corvus, but I believe that is still post gate | 13:29 |
| evrardjp | understood. | 13:30 |
| evrardjp | ok let's build 3 times. | 13:30 |
| evrardjp | thanks pabelanger ! | 13:31 |
| pabelanger | evrardjp: you can use some of the regional caches for docker to maybe help speed things up | 13:31 |
| *** rlandy has joined #zuul | 13:33 | |
| evrardjp | I am using them already | 13:38 |
| evrardjp | :) | 13:38 |
| evrardjp | using the install-docker role auto configures them | 13:38 |
| evrardjp | if that's what you meant | 13:38 |
| *** bhavikdbavishi has quit IRC | 13:44 | |
| *** jpena|lunch is now known as jpena | 13:48 | |
| *** nilashishc has joined #zuul | 14:25 | |
| *** hashar has quit IRC | 14:30 | |
| *** irclogbot_3 has quit IRC | 14:41 | |
| *** irclogbot_3 has joined #zuul | 14:51 | |
| *** smyers has quit IRC | 14:57 | |
| *** smyers has joined #zuul | 15:00 | |
| *** irclogbot_3 has quit IRC | 15:04 | |
| *** irclogbot_3 has joined #zuul | 15:16 | |
| *** irclogbot_3 has quit IRC | 15:20 | |
| *** irclogbot_3 has joined #zuul | 15:30 | |
| corvus | evrardjp: yeah, soon we'll be able to drop it to 2 builds check / gate+post. but check and gate will frequently be different; further reduction would be complex. | 15:30 |
| evrardjp | no that's fine I was just looking for gate+post | 15:33 |
| evrardjp | thanks corvus | 15:34 |
| *** swest has quit IRC | 16:20 | |
| *** swest has joined #zuul | 16:22 | |
| *** swest has quit IRC | 16:26 | |
| *** rfolco is now known as rfolco_doctor | 16:28 | |
| *** bhavikdbavishi has joined #zuul | 16:31 | |
| *** Alvass has joined #zuul | 16:34 | |
| Alvass | Hi, I might have found a bug with zuul-executor executor/server.py | 16:36 |
| *** themroc has quit IRC | 16:39 | |
| *** Alvass has quit IRC | 16:47 | |
| clarkb | tobiash: pabelanger dmsimard https://review.openstack.org/#/c/624525/4 aims to add repo mirror support to where we install ovs on centos for multi-node-bridge role. There is one comment in there I likely need to address but looking for input from outside of openstack for whether or not that change makes sense for you | 16:50 |
| clarkb | in aprticular we've found taht the centos mirrors are incredibly flaky particularly for centos-ceph-luminous | 16:50 |
| clarkb | and so want to be sure we are using our mirror when those are hit | 16:50 |
| *** electrofelix has quit IRC | 16:53 | |
| *** quiquell is now known as quiquell|off | 17:06 | |
| *** ianychoi has joined #zuul | 17:38 | |
| *** panda is now known as panda|off | 17:43 | |
| *** nilashishc has quit IRC | 18:05 | |
| *** rfolco_doctor is now known as rfolco | 18:12 | |
| *** jpena is now known as jpena|off | 18:17 | |
| *** bhavikdbavishi has quit IRC | 18:25 | |
| *** irclogbot_3 has quit IRC | 18:40 | |
| *** irclogbot_3 has joined #zuul | 18:44 | |
| dmsimard | clarkb, pabelanger: I was looking at ^ and noticed this: https://github.com/openstack-infra/zuul-jobs/blob/master/roles/multi-node-bridge/tasks/common.yaml#L45-L51 | 18:47 |
| dmsimard | That doesn't do what you think it does | 18:47 |
| dmsimard | centos-release-openstack-queens (there's rocky now too btw) installs two other packages: centos-release-qemu-ev and centos-release-ceph-luminous | 18:48 |
| dmsimard | Uninstalling centos-release-openstack-queens does *not* uninstall centos-release-qemu-ev and centos-release-ceph-luminous | 18:48 |
| dmsimard | Illustrated: http://paste.openstack.org/raw/737159/ | 18:50 |
| clarkb | dmsimard: si there a yum clean or similar we can run to remove those no longer required transitive deps? | 18:50 |
| dmsimard | clarkb: the dependency is in one direction (from "cloud" to "virt" and "ceph") that's why they aren't being removed | 18:51 |
| clarkb | dmsimard: ya with apt you can do apt-get autoremove iirc to clean up those dangling deps | 18:51 |
| dmsimard | I don't know if there's a way to "clean" it up other than to explicitely uninstall those repositories as well | 18:51 |
| dmsimard | In https://review.openstack.org/#/c/580518/ we changed from using rdo-release which was what I had originally implemented to using the centos-release package which has this problem | 18:52 |
| dmsimard | the rdo-release package is self-contained and ships only the RDO and Virtualization repositories -- more importantly, uninstalling rdo-release removes both repositories | 18:52 |
| clarkb | dmsimard: I'm the wrong person to try and untangle what all that means and decide on a proper fix. But I'm happy to help test/review other suggestions if therea re better methods | 18:53 |
| clarkb | dmsimard: to summarize what I'm trying to address is mirror.centos.org is super flaky (particularly for ceph-luminous) and I want to ensure we hit our region mirrors for ovs installs there | 18:53 |
| dmsimard | clarkb: https://review.openstack.org/#/c/580518/ mentions "Drop the dependency on RDO infra" which refers to the rdo-release package being hosted on rdoproject.org and we can't exactly claim it has 100% uptime | 18:54 |
| dmsimard | clarkb: sure, but ceph-luminous shouldn't be installed to begin with | 18:55 |
| dmsimard | that's why I went and looked at what was going on :D | 18:55 |
| clarkb | dmsimard: except that it is pulled in by the package that installs ovs | 18:55 |
| clarkb | dmsimard: its the metadata that fails not ceph packages | 18:55 |
| dmsimard | it's pulled in by centos-release-openstack-queens | 18:55 |
| clarkb | yup | 18:55 |
| dmsimard | Worst case scenario, we could probably even vendor the rdo-release rpm in the role | 18:56 |
| dmsimard | It has no dependencies and is self contained -- I mean, it's just gpg keys and repo files | 18:56 |
| clarkb | dmsimard: but that will still pull the ovs package from elsewhere too right? | 18:56 |
| dmsimard | clarkb: the location of the ovs package is already proxied by infra | 18:56 |
| clarkb | there are two levels of concern here, first is the repository setup package needing to be reliable the second is the ovs packages themselves needing to be reliable | 18:56 |
| clarkb | ah so we can point it at that location | 18:57 |
| dmsimard | hang on | 18:58 |
| clarkb | it is unfortunate that a completely unrelated package is so tightly coupled to "openstack" | 18:58 |
| dmsimard | +1 | 18:58 |
| dmsimard | clarkb: i.e, http://mirror.iad.rax.openstack.org/centos/7/cloud/x86_64/openstack-rocky/openvswitch-2.9.0-3.el7.x86_64.rpm | 19:01 |
| clarkb | ya thats the centos-openstack-release-rocky repo right? not the rdo one? | 19:01 |
| clarkb | whcih is what we are trying to use today (but s/rocky/queens/) | 19:01 |
| dmsimard | RDO ships packages in CentOS as part of the Cloud SIG | 19:02 |
| dmsimard | which is that repo | 19:02 |
| clarkb | right, but they have different repo locations depending on where you grab the file? | 19:02 |
| clarkb | if they are the same, then why does centos-openstack-release-queens dep on centos-ceph-luminous | 19:02 |
| dmsimard | no, it's just centos-release-openstack which installs extra things -- basically this: https://github.com/rdo-infra/centos-release-openstack/blob/rocky-rdo/centos-release-openstack.spec#L14-L15 | 19:03 |
| clarkb | in that case, maybe we do the vendory type thing. Just write out the .repo file and the gpg key (without pulling in ceph or anything else) and isntall ovs | 19:03 |
| clarkb | that should work assuming ovs doesn't dep on ceph (which it shouldn't) right? | 19:04 |
| dmsimard | it doesn't -- I attempt to install it in my pastebin above | 19:04 |
| dmsimard | it's pulling a package from base and extras, that's it | 19:04 |
| clarkb | cool, I like that approach then. We can set to upstream repo location if no mirror is set or use mirror if set | 19:04 |
| dmsimard | Considering that particular OVS package doesn't require any other dependencies | 19:05 |
| clarkb | and set it to disabled and explicitly enable it when we install ovs | 19:05 |
| dmsimard | You could even install it over http without setting up any repos | 19:05 |
| dmsimard | Like "yum install http://mirror.iad.rax.openstack.org/centos/7/cloud/x86_64/openstack-rocky/openvswitch-2.9.0-3.el7.x86_64.rpm" | 19:05 |
| clarkb | will that verify gpg? probably not. The verification is probably worth having | 19:05 |
| dmsimard | true, also requires you to update that url as versions come and go | 19:06 |
| dmsimard | I can send a patch -- do we want to vendor the rdo-release.rpm file ? Or just the gpg-key and the .repo file ? | 19:07 |
| dmsimard | probably better to have the gpg-key and .repo file -- rpm is binary | 19:08 |
| clarkb | I would just do the .repo and gpg file rather than vendoring a "binary" format (its tar/ar right?) | 19:08 |
| clarkb | yup | 19:08 |
| dmsimard | ok | 19:08 |
| clarkb | thank you! | 19:08 |
| dmsimard | np, wish it'd be more straightforward | 19:09 |
| *** hashar has joined #zuul | 19:18 | |
| clarkb | dmsimard: can you ping me when you get a patch up? I'll be sure to review it quickly. | 19:49 |
| *** kmalloc is now known as notmorgan | 20:10 | |
| *** notmorgan is now known as morgan | 20:10 | |
| dmsimard | clarkb: yup, will do. | 20:16 |
| *** hashar has quit IRC | 21:10 | |
| pabelanger | clarkb: corvus: are we in a position to tag a new release of nodepool / zuul this week? Was hoping to start trying out the executor zone stuff for ansible-network | 21:32 |
| clarkb | seems to be stable for infra | 21:44 |
| pabelanger | I haven't see, is the latest check queue stuff enabled yet? | 21:45 |
| clarkb | scheduler is running that codebut change to use it hasnt merged | 21:45 |
| openstackgerrit | David Moreau Simard proposed openstack-infra/zuul-jobs master: Vendor the RDO repository configuration for installing OVS https://review.openstack.org/624817 | 21:49 |
| dmsimard | clarkb: ^ something like this ? | 21:49 |
| dmsimard | pabelanger: ^ | 21:49 |
| clarkb | dmsimard: woot will review shortly | 21:50 |
| clarkb | dmsimard: just one question but I think that looks great (the tests should answer the question too) | 21:55 |
| dmsimard | clarkb: good point | 21:56 |
| openstackgerrit | David Moreau Simard proposed openstack-infra/zuul-jobs master: Vendor the RDO repository configuration for installing OVS https://review.openstack.org/624817 | 21:57 |
| clarkb | dmsimard: you can also update the depends on at https://review.openstack.org/#/c/624526/ to point at your change and exercise it a bit more on centos specifically via the tripleo jobs | 21:57 |
| dmsimard | clarkb: I did that | 21:58 |
| clarkb | great if the tests come back clean on both chagnes I think that is good indication its working fine and we can proceed to merge it | 21:59 |
| dmsimard | wait, I just did something stupid | 21:59 |
| dmsimard | ok, fixed -- I had 624526 depend on itself :D | 22:00 |
| clarkb | ha | 22:00 |
| *** kklimonda_ has joined #zuul | 22:23 | |
| *** smyers_ has joined #zuul | 22:23 | |
| *** maxamillion has quit IRC | 22:27 | |
| *** smyers has quit IRC | 22:29 | |
| *** kklimonda has quit IRC | 22:29 | |
| *** logan- has quit IRC | 22:29 | |
| *** smyers_ is now known as smyers | 22:29 | |
| *** logan- has joined #zuul | 22:31 | |
| *** gouthamr has quit IRC | 22:32 | |
| *** gouthamr has joined #zuul | 22:35 | |
| *** maxamillion_ has joined #zuul | 22:53 | |
| *** maxamillion_ has left #zuul | 22:53 | |
| *** maxamillion_ has joined #zuul | 22:53 | |
| *** maxamillion_ has left #zuul | 22:54 | |
| *** maxamillion_ has joined #zuul | 22:54 | |
| *** maxamillion_ is now known as maxamillion | 22:56 | |
| *** maxamillion has quit IRC | 22:56 | |
| *** maxamillion has joined #zuul | 22:56 | |
| openstackgerrit | Clark Boylan proposed openstack-infra/zuul-jobs master: Vendor the RDO repository configuration for installing OVS https://review.openstack.org/624817 | 23:14 |
| clarkb | ianw: ^ is that what you had in mind for dmsimard's change? (I'm pushing revision up to try and pseed up turnaround on that as its impacting a few jobs) | 23:14 |
| clarkb | I also fixed atemplating error that testing found | 23:14 |
| dmsimard | clarkb: thanks, I was driving back home from the office and got stuck in traffic. Eating dinner and I'll review | 23:16 |
| dmsimard | Oh man, I used "fi" instead of "endif" | 23:18 |
| dmsimard | Bash leaking into my jinja | 23:18 |
| ianw | clarkb: yep, i think having a non-conflicting name is good | 23:18 |
| dmsimard | ianw, clarkb: the name was non-conflicting to begin with, unless mistaken | 23:18 |
| *** rlandy is now known as rlandy|bbl | 23:28 | |
| ianw | dmsimard / clarkb: hrm, looks the same to me? I noticed we do similar with the gpg file | 23:48 |
| dmsimard | ianw: ah, agreed on the gpg file name, I did use the same | 23:58 |
| dmsimard | repo should not clash however | 23:58 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!