[ https://ovirt-jira.atlassian.net/browse/OVIRT-2201?page=com.atlassian.jir... ] Barak Korren commented on OVIRT-2201: ------------------------------------- Pasting IRC transcript of me trying to understand the user story behind this {code} <sbonazzo> bkorren: maybe there are better ways, but to me this is a regression that complicates my work experience and make me rethink about the switch to v2 <bkorren> sbonazzo, my point is, that v2 gave you certain ways to do things; but we shouldn't fixate on those, if we can examine the use case and figure out ways to deliver what you really need in a more convenient way <sbonazzo> bkorren: I need an end point in jenkins pointing me to latest successfull build per branch per project * ena has quit (Ping timeout: 622 seconds) <bkorren> sbonazzo, must it be in jenkins? if I put it in resources for example? <bkorren> sbonazzo, does it need to be the build with logs , etc, or just the artifacts? <sbonazzo> bkorren: resources may be ok too. artifacts only may be ok too, including yum repodata directory <sbonazzo> bkorren: but that will make difficult to correlate the artifact to the job <bkorren> sbonazzo, another question - will an HTML page with links be enough, or would you need predictable URLs? <bkorren> sbonazzo, that begs the question about why do you need to correlate the artifact with the job? is the job just a way to get at the patch? or do you need the build logs? <sbonazzo> bkorren: predictable urls would work far better for me. about correlation, it may be needed in the only case latest successful build results older than the packages in testing, requiring an inspection on what happened <bkorren> sbonazzo, I wonder about the root cause for all these requirements, is there a certain work process here we can revisit and modify so it becomes less labor intensive and requires less data gathering and manual correlation? <bkorren> sbonazzo, in other words - is there a way to "short-circuit" these requirements and make a system that will just deliver the end result you need instead of parts you need to gather manually? <sbonazzo> bkorren: there are cases when you merge a patch and the build keeps hanging in change queue for months before landing to tested due to some reason. I want to be aware of this happening in a shorter loop so i want to check that latest build lands in tested within the day. <sbonazzo> bkorren: I also would like to be able to have a quick look to a page (up to now it was a jenkins view) showing me that master was failing while 4.2 was fine and not the other way around <bkorren> sbonazzo, again if the point is to know what landed in 'tested' when - I'll just find one or more ways to deliver that exact info back to gerrit. <bkorren> sbonazzo, when you say 'master failing' here you mean builds for master branch of a certain project, if patches from master branch of said project are passing CQ/OST or if 'ovirt master' is OK from an OST POV? <sbonazzo> bkorren: a master jenkins job failing for a certain project. not sure you can see it, but have a look at: https://jenkins.ovirt.org/user/sbonazzo/my-views/view/Failing%20by%20rele... <bkorren> sbonazzo, what is the purpose of gathering this information? I see a whole mix of things - check patch mixed with build artifats and with check merged... <sbonazzo> bkorren: getting an insight on what's failing per target <sbonazzo> bkorren: since we are not gating the builds on check-merged I need to know if we built something which is not passing check-merged <bkorren> sbonazzo, adding gatting on check-merged to CQ would be quite a low hanging fruit from me if it means you can spend less time monitoring things manually... <bkorren> sbonazzo, I gate on build-artifacts already, actually now that I think of it - v2 already gives me that... <sbonazzo> bkorren: gating will help indeed :-) not sure how RDO managed to do it on their automation, but I really like zuul <bkorren> sbonazzo, I liked zool, until I fugure how rigid was it <bkorren> sbonazzo, but I think we're talking about different things - there is pre-merge gating and post-merge gating <bkorren> sbonazzo, pre-merge gating is VERY expensive to accomplish, and while CQ was designed with that in mind - I don't think its likely we'll go there at this point; but is we say that what you want to know is that a package or a patch that reached 'testing' had passed check-merged, then as I said, we already do that in V2. * sbonazzo is now known as sbonazzo|mtg <bkorren> sbonazzo, the issue with zuul is that it just assumes infinite hardware or very cheap tests... <sbonazzo|mtg> bkorren: so all v2 built packages landing in tested are certified passing check-merged? <bkorren> sbonazzo|mtg, yes. this is a side-effect of both check-merged and build-artifacts running within the same job as opposed to different jobs; I'm surprised this did not occur to me or I would've listed it as a benefit somewhere... :) <sbonazzo|mtg> bkorren: :-D {code} -- This message was sent by Atlassian Jira (v1001.0.0-SNAPSHOT#100087)