Monday, 2022-03-14

*** frenzyfriday|PTO is now known as frenzyfriday05:20
*** ysandeep|out is now known as ysandeep06:01
*** ysandeep is now known as ysandeep|lunch07:20
*** ykarel is now known as ykarel|LUNCH08:16
*** jpena|off is now known as jpena08:38
*** ysandeep|lunch is now known as ysandeep08:47
*** odyssey4me|away is now known as odyssey4me09:00
*** odyssey4me is now known as odyssey4me|away09:00
*** odyssey4me|away is now known as odyssey4me09:02
*** odyssey4me is now known as odyssey4me|away09:25
*** odyssey4me|away is now known as odyssey4me09:25
*** ykarel|LUNCH is now known as ykarel09:36
*** arxcruz is now known as arxcruz|lunch11:04
*** dviroel is now known as dviroel|ruck11:14
*** arxcruz|lunch is now known as arxcruz12:04
*** odyssey4me is now known as odyssey4me|away12:18
*** odyssey4me|away is now known as odyssey4me12:18
*** odyssey4me is now known as odyssey4me|away12:33
fricklerodyssey4me|away: any chance you could stop this |away nick toggling?12:40
*** odyssey4me|away is now known as odyssey4me12:45
*** odyssey4me is now known as odyssey4me|away12:46
*** odyssey4me|away is now known as odyssey4me12:50
*** odyssey4me is now known as odyssey4me|away12:50
*** odyssey4me|away is now known as odyssey4me13:00
*** ysandeep is now known as ysandeep|break13:20
odyssey4mefrickler my bad, lemme figure out what's doing that13:35
*** ysandeep|break is now known as ysandeep13:58
*** pojadhav is now known as pojadhav|afk15:08
*** marios is now known as marios|call15:31
*** dviroel|ruck is now known as dviroel|ruck|lunch15:36
clarkbanything important I should be catching up on today/this week now that I am back?15:41
clarkbLooks like one of the backup servers needs pruning15:43
fungiyeah, i haven't gotten to that yet. i was also going to switch our openstackid ssl check entry over to openinfraid now that we don't run anything depending on the former15:48
fungiand there's a dns change requested for the ci-log-processing systems15:49
fungii'll get to that in a bit15:49
fungithere was some disruption over the weekend where boot attempts in ovh-bhs1 were consistently failing. i haven't looked to see if that's still occurring15:49
fungiand we got some maintenance notices in french about gra1 api availability for 19:00-01:00 tomorrow (starting roughly 27 hours from now) but i'm inclined to just let nodepool complain15:50
clarkb++ to letting nodepool complain15:58
clarkbsounds like I can mostly catch up via reviews then. I'll try to start there15:59
*** marios|call is now known as marios16:02
clarkbfungi: grafana shows we have in use nodes in ovh bhs116:03
clarkbI suspect nodepool is happy there again16:04
clarkbI'd like to make progress on the gitea 1.16 upgrade this week, but have noticed that we're very much into openstack's release period and not disrupting git hosting for a bit may be a good idea. Do others have opinions on that?16:04
clarkbif it was a 1.15.x update I think I'd be fine with it16:05
*** ysandeep is now known as ysandeep|out16:12
*** melwitt_ is now known as melwitt16:14
*** dviroel|ruck|lunch is now known as dviroel|ruck16:22
mnasiadkagood afternoon, any DIB core can help merging ?16:25
clarkbmnasiadka: thats the first time I've seen : used to assign a default value to a var. I'm a bit confused on how that works since : is supposed to be a noop?16:29
clarkbI guess I'm used to FOO=${FOO:=foo}16:30
clarkboh I see its the := that is important here16:30
clarkband then : chumps the substitution16:30
mnasiadkayes, that's what I found out googling over the mighty internet16:31
clarkbmnasiadka: I'm just used to the construct in devstack whcih is the other one. Both seem to work and I understand this one now :)16:32
mnasiadkaclarkb: fine by me, it seems you can learn something new every day :)16:32
fungiclarkb: i doubt gitea upgrades will be all that disruptive so long as we're testing its basic functionality. it's gerrit upgrades i'd be more concerned with at this time16:33
mnasiadkaianw: can you create a release after merges?16:37
clarkbfungi: ok, I don't think that is a today item, but maybe midweek I'll have time to land that and cross it off my list. For today trying to catch up on reviews and prepare for the meeting tomorrow etc etc16:40
fungisounds great!16:46
*** marios is now known as marios|out16:54
*** jpena is now known as jpena|off17:35
fungiheaded out to an appointment, should be back in an hour-ish17:41
*** arxcruz is now known as arxcruz|off17:54
opendevreviewMerged openstack/diskimage-builder master: containerfile: add support for Docker
opendevreviewNeil Hanlon proposed opendev/base-jobs master: Add rockylinux nodeset
opendevreviewMerged openstack/diskimage-builder master: Handle btrfs root subvolume for fedora extract-image
iurygregorydiablo_rojo_phone, quick question regarding PTG rooms, by any chance there a chance to have a new room on Tuesday so the ironic team can have a session from 14 to 17 UTC? (not sure if this would be the right channel to ask you this question =) )19:45
diablo_rojo_phoneiurygregory: Like we add a Newton or wherever in the alphabet we are in that ethercalc because the rest of that time window is full? 19:48
iurygregorydiablo_rojo_phone, yup =)19:51
diablo_rojo_phoneI thiiiiink we can make that work, but I will double check with the powers that be and have to get back to you. 19:51
iurygregorydiablo_rojo_phone, sure, thank you for checking! it would be only on Tuesday April 519:52
clarkbThis reminds me we should do more testing of meetpad and the sound issues we've observed.19:52
* clarkb made a note to try and debug it more this week19:52
fungihappy to help test19:55
funginow that i've worked out the nuances of sound on my workstation19:55
clarkbfungi: cool maybe tomorrow morning?19:56
clarkbmostly just need to see if we reliably get working audio when connecting from various browsers with sound pre muted and pre unmuted in pavucontrol19:57
fungii've got stuff on my schedule from 13:30 to 16:00, but open between then and the opendev meeting19:58
clarkbsounds good19:58
fungii guess that's 9am your time, so morning enough hopefully19:58
clarkbyup thats perfect actually20:01
corvus is the fix in zuul for full-reconfigure not correcting the branch cache after the missed event.  it doesn't seem like a high priority right now since frickler re-created the one branch we saw a problem with.  that change should help the next time it happens though.20:03
diablo_rojo_phoneclarkb: yeah let me know if you need help testing. I can try to help too since I have the app on my phone now etc. 20:04
fungithanks corvus!20:04
ianwmnasiadka: yes, can do a release20:08
ianwi'll make it 3.20 because we have this change to containerfile, and a bigger one for the boot/efi partitions in incoming images20:27
*** dviroel|ruck is now known as dviroel|ruck|afjk20:50
*** dviroel|ruck|afjk is now known as dviroel|ruck|afk20:50
ianwcorvus: as someone recently playing with dashboards, series may be of interest.  we've had the json import for a while, but this is the first time i've really attempted to use it.  editing locally via the UI is quite good20:58
ianwthey have deprecated the old graph panel time for a new "time series" panel, which grafyaml doesn't have the schema for20:59
ianwit's of course very different data structures; afaics it's not really defined anywhere, apart from the actual ts files.  reverse engineering it isn't really on my todo list21:00
ianwi think they keep it quite backwards compatible, so you can import old dashboards.  but i don't think they explicitly export the panel types as an api, as such21:01
mnaserhi folks -- i'm wondering if it's okay to publish non-released versions of things to -- mainly thinking/looking of having an ansible collection build of each commit published there (or is there a better 'scratch' space location?)21:07
mnaserit's just a bunch of yaml file(tm) so i cannot imagine it taking up much space21:07
clarkbmnaser: we have asked that you not publish a new file for each commit but rather update a single file in the past21:08
clarkbeg the master tarball gets updated each time master updates21:08
mnaseroh that's also a perfectly reasonable request21:08
mnaserthat covers the "get the latest package" which is really what is wanted here21:08
clarkbthere should be examples of that for openstack python sdists21:09
mnaserit's an ansible collection, so i think we can just build it using the pbr version, and then after that the file with the same name21:10
mnaserokay, that sounds pretty reasonable, i'll work off of that then21:10
corvusianw: the value i get from grafyaml isn't the schema validation, it's the yaml->json conversion... can we just drop the schema validation and keep storing our panels in yaml?  i don't really consider json to be human-editable, so i don't feel like i can edit or create a new graph easily with json.  but that's exactly what i do with the yaml.21:14
corvusianw: when i made the zuul performance metrics dashboard, i did develop it with local grafana installation, but only enough to get one graph the way i wanted, then i used grafyaml to make 30 similar graphs with different metrics.  that was a lot more convenient than the alternatives.21:19
mnaseris there a reason why this job doesn't have the secret attached to it directly?
clarkbmnaser: the docstring points to it. Each tenant will have different secrets21:43
mnaserthe secrets that are attached to it is the same, i'm looking to publish an ansible collection (.tar.gz file) and the options aren't really usable21:43
clarkbthey should be different per tenant iirc. Eg zuul's is different than openstacks.21:44
mnaserah, we've been mostly using/relying on opendev infra (so publish to, etc)21:44
ianwcorvus: it's not written that way, but it could be.  honestly if you're developing the graph using the ui, then there might be better ways to have similar graphs, such as template variables21:44
clarkbya zuul publishes to a zuul specific dir in afs and openstack to an openstack specific dir. I think if you use the opendev root then ya it can be shared?21:44
clarkbbut that isn't true for all tenants/usages21:44
mnaseryeah my thought process didnt think of that part21:45
corvusianw: i did use template variables; that serves a very different function than graph repitition.21:46
mnaseri guess i can use the `opendev-promote-python` job as a parent and then override the vars..21:46
corvusianw: my point was that most of what grafyaml does is translating yaml to json, and i think that's valuable.  we obviously could have just stored the graphs as json to start with, but we chose not to, and we chose to implement grafyaml instead to make it easier to maintain21:47
corvusianw: i still feel that grafyaml *does* make it easier to maintain than just checking in json blobs.21:47
corvusianw: so, as an opendev contributor, that's my feedback: grafyaml makes that work easier to do, and i think it's worth keeping the graphs in yaml for that purpose.  my recent dashboard contribution was directly facilitated by that, and i do not think that i would have made the same contribution without it because the alternative is so tedious and error-prone.21:51
clarkbcorvus: ianw: wasn't one of the ideas that you could build the dashboard in grafana then export the json? So you aren't editing json directly, but you also arent' editing yaml and instuead clicking in a browser which may also not be desireable?21:55
ianwok.  i think that telling a new contributor -- no you edit it using the GUI locally, but then you open up the json, and you figure out using undocumented rules, mostly cargo-cult copying from previous graphs, what are the important bits to extract into YAML, then put it in the tool -- which will schema validate some of it, and turn it back into json, that we will re-import ... is not an encouraging workflow.21:55
corvusclarkb: well, that's what ianw is proposing we do -- basically just use grafyaml as a json importer.  i'm saying that json is not human-editable nor is it human-reviewable, so from my perspective, we're just checking in blobs.21:56
ianwclarkb: yes, that is my idea, in a nutshell, would be the first example of it21:56
corvusianw: we've had great success with new contributors adding new dashboards with grafyaml.  what you describe is not at all the actual process.21:57
clarkbcorvus: got it, I think that was the context I was missing21:58
corvus(except the cargo-cult part -- that is, very much, intentionally the process, and that's the best thing about it)21:58
opendevreviewClark Boylan proposed opendev/system-config master: Update Gitea to 1.16.4
opendevreviewClark Boylan proposed opendev/system-config master: DNM change to test and hold gitea 1.16
ianwexcept if you want anything that grafana has added outside what is already cargo-culted, you have to reverse engineer the undocumented grafana json behind it21:59
clarkbI've put an autohold in place for to grab the gitea and we can check it22:00
clarkbianw: is the yaml grafyaml accepts a 1:1 yamlification of the json that grafana outputs?22:01
clarkbianw: I wonder if we can send it through ruamel or similar22:01
corvusyeah, that's the unfortunate part.  but that is what we decided to do when we wrote grafyaml, and that's the investment that makes it worthwhile.  it sounds like there is a new feature in grafana you want to use that would require some more investment in grafyaml and you don't want to make that investment.  that's understandable, but let's be clear about what you're suggesting: we stop using yaml altogether (essentially only using grafyaml as a22:02
corvusjson blob importer).22:02
corvusrather than stop using grafyaml, i would rather we 1) invest in adding support to grafyaml for the new feature, or 2) don't use the new feature.  i think the benefits grafyaml bring are that important.22:02
clarkbcorvus: what is the new feature?22:03
corvusclarkb: grafyaml is an extremely thin yaml->json converter (it is so thin it looks like it does almost nothing, but it does enough that apparently the new feature ianw wants to use requires changes)22:03
corvusclarkb: i don't know22:04
clarkbgot it22:04
ianwgrafana has changed the default to a "time series" panel22:05
ianwgrafyaml does not have the schema validation for this22:05
ianwyou'll get a warning in the UI that the old panel type is deprecated, and they have a button to "upgrade" it to the new panel type22:06
clarkbI guess that means we need to update grafyaml to output the new type from the existing yaml inputs? and I guess validate that this is valid22:06
ianwwell, yes, i guess that would be the equivalent of writing the upgrade tool that grafana provides that moves graph panels -> timeseries panels22:07
corvusor change grafyaml to grok the timeseries schema and then "upgrade" our yaml by hand22:08
corvus(i think that makes more sense than encoding a translation layer in grafyaml; i like it being a thin layer)22:08
*** rlandy is now known as rlandy|bbl22:14
mnaserok cool, so going to use zuul artifacts to publish per-commit collections (which i dont care about being long lived), and then i will take those and promote them later into pointing to specific branch22:18
clarkbmnaser: ya the zuul artifacts will live for 30 days if written to the log servers (which most artifacts are though not all)22:19
mnaseryeah, it's good enough for the use case of "i want to try pulling this recent build"22:20
mnaserand then the published to tarballs.o.o will be the static "gimme latest"22:20
ianwit's not just the time series, though. you can see everything goes in a "fieldConfig" dict now.  the closest i've found to documentaiton of what this is, is
ianwwhich is pretty much just fairly useless autogenerated fluff22:24
ianwthat has a "custom" field where it seems actual things that affect the graph display are put.  22:25
corvusif the json is not comprehensible then we can't review it either.22:27
clarkbis it possible to take the json from a change and put it into a grafana and validate it that way? I suppose that doesn't mitigate against risks if there is some sort of flaw in the input though22:33
ianwI feel a better way to review than having people try to be human grafana undocumented api converters would probably be to pull the change, and start up a grafana using something like and view the dashboard in-situ.  at this point we could even have CI do screenshots 22:33
ianwbut i understand you have a fundamentally different view, if we don't want to support both paths, then ok, we have what we have until we figure out how to update grafyaml22:34
clarkbya I think that could verify it generates graphs we like, but not that there isn't some hidden nefarious functionality. Though that may not be a possibility with grafana at all22:34
ianwthe json really isn't *that* bad to parse. is a concrete example22:35
ianwanyway, i'm out afk for now22:41
clarkbfungi: on the project retirement side of things we retired all the repos and abandoned their changes right? We should be good to land now?22:59
clarkbinfra-root if you can double check that there isn't anything else to add to the meeting agenda I'll get that sent out soon23:03
fungiclarkb: yes, for all the retirement changes set to your chosen topic, i approved them all and then abandoned every other open change for those same repos23:11
clarkbfungi: excellent then I think that change is ready23:13
fungii concur23:14
clarkbI'm going to send out the agenda now. We can always edit it later if something important comes up23:15
clarkbI remembered to add a note about the DST change in my email too :)23:17
clarkbhttps:// here is the held gitea server. Maybe we can land that upgrade tomorrow? Reviews and double checking with the test node very much welcome23:20
clarkbAt first glance it seems fine23:20
*** dviroel|ruck|afk is now known as dviroel|ruck23:49

Generated by 2.17.3 by Marius Gedminas - find it at!