Friday, 2026-04-24

*** vhari_ is now known as vhari03:12
opendevreviewFrancois Magimel proposed openstack/openstack-zuul-jobs master: fix: get the module name from pyproject.toml  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/98603707:17
opendevreviewFrancois Magimel proposed openstack/openstack-zuul-jobs master: fix: get the module name from pyproject.toml  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/98603707:21
*** ykarel_ is now known as ykarel14:35
cardoehttps://review.opendev.org/dashboard/?title=Open+Reviews+for+Ironic&foreach=is:open+-is:wip&ironic-week-prio=(hashtag:ironic-week-prio+status:open)&Ironic=((project:%5Eopenstack/ironic.*)+OR+project:openstack/sushy+OR+project:openstack/python-ironicclient)&Emulators=(project:openstack/sushy-tools+OR+project:openstack/virtualpdu+OR+project:openstack/virtualbmc)&Networking=(project:openstack/networking-generic-switch+OR+projec15:42
cardoet:openstack/networking-baremetal)&Tooling=(project:openstack/bifrost+OR+project:openstack/metalsmith+OR+project:openstack/tenks+OR+project:openstack/molteniron)15:42
cardoeI guess my query is a bit long... but I'm getting a 414 opening that URL from the /changes/ endpoint.. wondering if there's some tweaks folks know about that I can do.15:42
fungihuh, i wonder if that's a behavior change in mina-httpd since the gerrit 3.12 upgrade?15:45
clarkbit could be coming from apache as well. Would need to look at logs15:46
fungithe "Endpoint: /changes/" part makes me suspect it's coming from deeper in15:47
JayFmy giant terrible query still works fwiw15:48
JayFhttps://review.opendev.org/q/status:open+and+((project:%5Eopenstack/ironi.*)+or+(project:%5Eopenstack/sush.*)+or+(project:openstack/virtualpdu)+or+(project:openstack/virtualbmc)+or+(project:openstack/metalsmith)+or+(project:openstack/molteniron)+or+(project:openstack/networking-baremetal)+or+(project:openstack/networking-generic-switch)+or+(project:%5Eopenstack/python-ironi.*)+or+(project:openstack/tenks))+and+is:mergeable+and+15:48
JayF-reviewedby:self+and+-owner:self15:48
cardoeI can screen shot it.15:48
JayFcardoe: just sanity check: you logged in? Not being logged in breaks a lot of queries15:48
cardoeJayF: yeah I'm trying to mutate that query.15:48
cardoeI am logged in.15:48
fungiJayF: your url is shorter than cardoe's, so that could be coming in just under the limit or something15:49
JayFoh cardoe if you're making my URL longer that's the entire problem15:49
JayFthat one is crafted to sneak under the URL length limit15:49
JayFit's why it doesn't cover 100% of Ironic projects15:50
fricklerit does say so explicitly in the error popup: The requested URL's length exceeds the capacity limit for this server.15:50
cardoeYep but the error endpoint is /changes/ which is what's odd to me15:50
cardoeBut I guess if you fought the limit Jay I'll abandon this.15:50
frickleriiuc it is not your url directly that is too long, but the query to the /changes/ endpoint that the UI generates from it15:51
JayFah, interesting15:51
cardoeTrying to use dashboards to make it possible to operate on multiple ones at the same time.15:51
cardoeEspecially with our push to make smaller commits.15:51
fungiif it's apache we could potentially look at adjusting https://httpd.apache.org/docs/current/mod/core.html#limitrequestline15:51
fungiseems to default to 8190 bytes15:52
fungiand cardoe's entire url is only 589 bytes15:53
fricklerfungi: if you look for the 414 in the gerrit access log, the resulting query to the /changes/ endpoint has 8224 bytes, so that's a bit above the 819016:08
fungiaha! so gerrit is internally looping back out to apache?16:08
fungirather than talking directly to itself?16:09
frickleriiuc it is the JS running in the browser that does this query16:09
cardoeholy that's a big explosion 589 bytes -> 822416:09
frickleralso it seems to list a large number of individual changes, so likely the number of changes matches triggers this16:10
fungioh! got it, the url that goes to gerrit initially returns a javascript payload for a dashboard that then remotely queries the actual api (through apache)16:11
clarkbya the js in your browser is makign subsequent requests which then go back through apache16:11
cardoeanother thing I've wanted to ask if we can consider (and maybe its per project) but https://gerrit-review.googlesource.com/Documentation/config-labels.html#label_copyCondition NO_CODE_CHANGE16:11
fungiso yeah, i guess maybe we could adjust limitrequestline for that16:11
cardoeI think that'll reduce some of the load on the test infra.16:11
cardoeLike I'll tell people to add "Closes-Bug: #XXXX" in the commit and now all the tests have to run again.16:12
clarkbcardoe: we already set it to copyCondition = changekind:NO_CODE_CHANGE OR changekind:TRIVIAL_REBASE in most cases16:12
fungicardoe: that's up to openstack to decide, but the argument against it in the past was that maintainers also review commit messages16:12
clarkbfungi: and openstack uses commit messages for semantic stuff like versioning16:12
fricklercardoe: maybe you want to add limits (on the number of changes) to your queries, cf. what kolla has https://tiny.cc/kolla-dashboard16:12
fungiright, that to, which can then affect job results16:13
cardoecommit messages change versioning? How?16:13
clarkbcardoe: pbr inspects commit tags and will bump generated versions appropriately based on those tags. Like new feature means a new feature version bump. Incompatible update means major version bump. I'm not a fan of the feature but openstack continues to use it aiui16:14
clarkb(not a fan because it creates confusion occasionally)16:14
fungicardoe: semver trailers16:15
cardoeokay interesting didn't know that existed.16:15
fungihttps://docs.openstack.org/pbr/latest/user/features.html#version16:16
cardoeI just had a patch that I ended up creating a launchpad item for nova and then trying to add the bug into the commit message and ended up needing to recheck 3 or 4 times with that being the only change.16:16
fungithings like `Sem-Ver: api-break`16:16
fungineeding to recheck the change 3 or 4 times suggests something is unhappy with testing, that may be worth looking closer into16:17
fungiavoiding running tests doesn't really help resolve the stability of test results16:17
cardoeI don't disagree.16:18
cardoeBeen trying to land a change in that project for a few years now so I'm not very motivated to research their tests.16:19
cardoeMaybe we can use parent projects? So that the query can be smaller?17:12
clarkbif the issue is in apache we could potentially increase that limit there too17:16
clarkbApache tends to make that stuff configurable. Not sure about mina/gerrit17:16
clarkbbut ya changing the representation in Gerrit maybe another approach that would be successfuly17:16
fungiclarkb: yeah, frickler confirmed that it's apache hitting the 8190 byte default request limit, so we should be able to tune that up from the default if we want17:18
cardoeI'm trying to make some dashboards for different teams to see patches more clearly.17:22
cardoeLike the kolla one you linked.17:22
cardoeSome cores on some projects tell me "I can never find changes that I need to review" which is apparently why some of my stuff has fallen in the cracks.17:23
cardoeSo projects are doing different things like an etherpad of patch links17:23
cardoeAnd it feels like I should be able to help them craft some queries to use gerrit17:23
cardoeSo not necessarily asking you folks to activate on anything. I'm just experimenting.17:27
clarkbyup, I guess let us know if you think something like the apaache configuration needs updating17:28
clarkbI think we'd be open to trying that. Worst case we dial it back down again if it causes unexpected load somewhere17:29
fungicardoe: also just in case you aren't aware, there's https://pypi.org/project/gerrit-dash-creator/ for storing and building them (i have no idea if it still works as i've never used it)17:34
cardoeah interesting17:36
fungiwith tooling developed by past generations of openstack maintainers, sometimes that lore doesn't get passed to subsequent generations17:37
clarkbalso gertty isn't perfect but I think exists to address many of these issues17:38
fungiyes, gertty does have a concept of dashboards too17:43
fungithough fundamentally they're just stored change query strings17:43
clarkbits tracking of changes that you haev reviewed and haven't reviewed and those that need to be rereviewed is a lot ebtter than the your turn thing in gerrit17:49
fungioh that too, yes17:51
cardoeThe your turn thing in gerrit is about as accurate as <brain too low from PTG>18:10
fungi...microsoft windows progress bars?18:14
clarkbI think it would be really neat to look at Gerrit's work organization stuff if I had time and really wanted to dig into Java. I think that there are a number of things that could be made better there. Like with submit requirements the system now knows when a change is missing one more review before it can merge. Automatically surfacing those would be great. Using heuristics18:30
clarkbto automatically do owner file type suggestions based on what you've already reviewed in the past and current changes in similar situation would be cool too18:30
clarkbI see what they are trying to accomplish with their rigid state tracking but I think it is too rigid and if one person doesn't follow the workflow they expect you to then the value approaches 018:30
clarkbbut if they tried to tackle the problem from a "good enough" perspective I think they have sufficient info to make suggestion pages for individual users based on the work they've already done and the state of changes against their submit requirements18:31
clarkb(the remaining sql db table captures reviewed by info on a per user repo file basis which is what I'd use to make those suggestions. Look at the last say month of files and repos reviewed by someone then suggest open  changes that touch those same files with an optional filter to see things that only need one more review to cross the finish line)18:34

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!