17:00:37 <jaypipes> #startmeeting
17:00:38 <openstack> Meeting started Thu Mar  1 17:00:37 2012 UTC.  The chair is jaypipes. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:00:39 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic.
17:00:43 <Ravikumar_hp> good morning Jay
17:00:57 <jaypipes> Ravikumar_hp: morning :)
17:01:05 <dwalleck> morning
17:01:06 <JoseSwiftQA> :)
17:01:23 <jaypipes> so... a quick status report from David Kranz and myself...
17:01:37 <jaypipes> #topic status report on stable/diablo Tempest branch
17:01:56 <jaypipes> So, the stable/diablo branch of Tempest was created and is now under Gerrit's control
17:02:10 <donaldngo_hp> nice!
17:02:16 <jaypipes> David and I have been working on fixes that allow Tempest to run against a Diablo environment smoothly.
17:02:19 <Ravikumar_hp> great . it helps
17:02:33 <jaypipes> Two fixes are already proposed against the branch:
17:02:45 <jaypipes> #link https://review.openstack.org/#q,status:open+project:openstack/tempest,n,z
17:03:01 <jaypipes> if you look at the above link, you will see the branch listed in the Gerrit output
17:03:15 <jaypipes> you can see there the two branches proposed for stable/diablo
17:03:21 <jaypipes> reviews very welcome.
17:03:37 <jaypipes> The big thing to point out is the following:
17:04:00 <jaypipes> a) Tempest's test cases will definitely be different from the stable/diablo branch to the development branch
17:04:38 <jaypipes> b) We had previously accepted a patch to development that added a "release" variable to tests that could be used to skip certain tests that failed on a particular release
17:05:14 <jaypipes> c) That skipping strategy should no longer be used. Instead, simply apply your patch to a specific branch of Tempest that is built to run against a specific release of OpenStack
17:05:29 <dwalleck> sounds like a much easier solution to me
17:05:39 <jaypipes> Any questions on the above? It's any important distinction and a big step towards aligning with other core projects
17:05:48 <Ravikumar_hp> jaypipes: question: where do we check-in common tests applicable for both ?
17:06:04 <jaypipes> Ravikumar_hp: excellent question! :) glad you asked!
17:06:12 <donaldngo_hp> so what this means is that if a bug is found in Essex Tempest then we need to see if needs to go into stable/Diablo
17:06:32 <jaypipes> donaldngo_hp: one sec, lemme answer ravi first... then I get to that one.
17:07:06 <jaypipes> OK, so now that we're set up correctly in Gerrit, here is the process for adding a test that is common to multiple releases:
17:07:21 <jaypipes> (and this is the same process for all core projects, so good to keep this in mind)
17:07:48 <jaypipes> 1) Always first propose a patch with your new test case to the development trunk (master)
17:08:09 <jaypipes> 2) Once that patch is approved, then propose the same patch against stable/diablo
17:08:33 <jaypipes> 2a) if any merge conflicts occur, of course, resolve those before proposing (or Gerrit will yell at you anyway :)
17:08:48 <jaypipes> So, the flow is always master -> stable branches
17:08:56 <Ravikumar_hp> thanks
17:09:04 <donaldngo_hp> jay there is likely hood that the patch will be for the same test but differnt assertions
17:09:15 <jaypipes> donaldngo_hp: precisely correct.
17:09:32 <donaldngo_hp> for example test_images.py minRam(Essex) and min_ram(Diablo)
17:10:17 <jaypipes> donaldngo_hp: correct. and that is exactly where you would change the test in the patch to stable/diablo to use min_ram instead of minRam
17:10:51 <donaldngo_hp> i think what that means is 1) run you proposed patch on DevStack(Essex) and then run it on DevStack(Diablo)
17:11:03 <donaldngo_hp> ok got it
17:11:04 <jaypipes> donaldngo_hp: this prevents all the if/else blocks on release variables... just apply the patch to correct release branch that makes sense for the API calls that release uses..
17:11:23 <donaldngo_hp> yep totally agree
17:11:23 <dwalleck> Is there a way to force Devstack to create a Diablo instance? My concern is that since I'm not primarily testing against diablo, it's much easier for me to propose tests that may fail
17:11:38 <jaypipes> dwalleck: another great question! :)
17:11:44 <jaypipes> dwalleck: indeed there is :)
17:11:58 <jaypipes> dwalleck: let me demonstrate:
17:12:12 <jaypipes> here are the steps I do when testing diablo locally
17:12:43 <jaypipes> $> rm -rf /opt/stack &&  cd $devstack_dir && git checkout gerrit/stable/diablo && ./stack.sh
17:13:07 <jaypipes> takes a while (since it needs to re-pull the branches), but it's guaranteed to be a clean diablo install\
17:13:32 <dwalleck> excellent! I can certainly do that
17:13:47 <donaldngo_hp> jaypipes: how do we shutdown DevStack I usually reboot my machine
17:13:51 <jaypipes> OK, so the end goal here is, of course, to have diablo clusters tested autmatically through jenkins against a diablo Tempest (and of course, same for essex)
17:13:58 <jaypipes> donaldngo_hp: sudo killall screen
17:14:07 <donaldngo_hp> sweet
17:14:24 <dwalleck> and now I don't feel so bad for doing the same thing :)
17:14:32 <jaypipes> donaldngo_hp: there's actually a patch coming (to master devstack) that has a restart.sh script... keep an eye out for that
17:14:45 <jaypipes> dwalleck: lol, hey, a reboot also works! :)
17:14:49 <donaldngo_hp> dwalleck++
17:14:57 <jaypipes> but sudo killall screen is much, much faster ;)
17:14:59 <Ravikumar_hp> we need also shutdown.sh
17:15:28 <jaypipes> Ravikumar_hp: echo "sudo killall screen" > shutdown.sh ;)
17:15:59 <Ravikumar_hp> ok
17:16:26 <jaypipes> OK, let's change topic to discussion of individual merge proposals and bugs. anyone object?
17:16:37 <Ravikumar_hp> jaypipes: we also need directories so as to group services
17:16:49 <jaypipes> Ravikumar_hp: could you elaborate?
17:16:50 <Ravikumar_hp> like nova , keystone, swift
17:17:07 <Ravikumar_hp> tempest/tempest/tests/nova ...
17:17:08 <jaypipes> Ravikumar_hp: talking about tempest or devstack?
17:17:12 <jaypipes> gotcha...
17:17:29 <Ravikumar_hp> we are planning to add keystone tests
17:17:39 <jaypipes> Ravikumar_hp: well, when I added the Glance tests, I put them all under tempest/tests/images/
17:17:45 <dwalleck> I agree. I think I have a merge prop with a compute test subdirectory
17:18:00 <jaypipes> Ravikumar_hp: so, when you add keystone tests, I would advise: tempest/tests/identity/
17:18:08 <Ravikumar_hp> sounds good
17:18:08 <dwalleck> I just didn't want to move everything since it didn't have to do with that bug
17:18:18 <jaypipes> anyone disagree with using the generic names in the tests directory?
17:18:34 <dwalleck> Vara is also about to submit some keystone tests as well, need to check in with him
17:18:42 <jaypipes> I did that to emphasize the tests are against the *API*, not the implementation...
17:18:51 <dwalleck> no, that sounds like it would be for the best
17:19:09 <jaypipes> dwalleck, Ravikumar_hp: OK, please coordinate with each other on keystone tests...
17:19:24 <Ravikumar_hp> sure . vara is ...?
17:19:46 <jaypipes> dwalleck: ^^
17:19:51 <dwalleck> Vara is another automation lead with Rack. He was going to be here but I don't see him...
17:20:01 <donaldngo_hp> i have a quesiton about: https://bugs.launchpad.net/tempest/+bug/943092
17:20:03 <uvirtbot`> Launchpad bug 943092 in tempest "test_servers_negative.py: Name error, release not defined" [Undecided,Fix committed]
17:20:15 <jaypipes> donaldngo_hp: that will be going away...
17:20:23 <jaypipes> donaldngo_hp: the release variable...
17:20:30 <donaldngo_hp> this was found by Sarad on my team and fixed
17:20:44 <donaldngo_hp> question is how did it ever pass the Jenkins build
17:20:48 <donaldngo_hp> should have failed?
17:20:56 <jaypipes> donaldngo_hp: because the jenkins built doesn't run temnpest :(
17:21:04 <donaldngo_hp> whoa
17:21:09 <donaldngo_hp> what does it do?
17:21:40 <jaypipes> donaldngo_hp: yeah... this is why we've been trying to get to a place where jenkins can consistently run tempest against the deployment cluster that currently runs devstacjk's exercises against it
17:21:59 <jaypipes> donaldngo_hp: but tempest has not been stable enough up to date
17:22:17 <donaldngo_hp> https://jenkins.openstack.org/view/Tempest/
17:22:28 <jaypipes> donaldngo_hp: and since there aren't any unit tests in tempest... there's not much to run in jenkins other than a pep8 checker and a merge conflicty check :(
17:22:38 <donaldngo_hp> i see
17:22:50 <jaypipes> donaldngo_hp: believe me, I'm ashamed about it...
17:22:57 <donaldngo_hp> was scratching my head wondering why its been green
17:23:14 <donaldngo_hp> makes sense now
17:23:16 <jaypipes> donaldngo_hp: and the breakout of stable/diablo was a step towards being able to run Tempest in a stable manner
17:23:34 <jaypipes> donaldngo_hp: that has always been our end-goal... to replace devstack's exercises with tempest.
17:23:47 <jaypipes> donaldngo_hp: and we're much closer today than we were a month ago atr least :)
17:23:53 <jaypipes> but, still lots to do!
17:24:10 <donaldngo_hp> awesome
17:24:21 <jaypipes> OK, shall we go over the merge proposals individually?
17:24:27 <jaypipes> #topic merge proposal status
17:24:34 <jaypipes> #link https://review.openstack.org/#q,status:open+project:openstack/tempest,n,z
17:24:40 <jaypipes> Let's go bottom to top...\
17:25:07 <jaypipes> the stress tests are going to be reviewed today by me. dwalleck and others, would be great to get a review from you!
17:25:40 <dwalleck> jaypipes: I'd like to. I just need to look clearly to try to understand how to run them
17:26:11 <jaypipes> Ravikumar_hp: it would be great to get those three Volume patches in. Could you focus on reviewing those? And ping Sapan about the comments on his review?
17:26:45 <Ravikumar_hp> we have volume & volume attachement - should we combine and make it as one?
17:27:00 <jaypipes> Ravikumar_hp: no... just saying to focus on those reviews..
17:27:12 <Ravikumar_hp> I will review and run it today
17:27:13 <jaypipes> Ravikumar_hp: they are different bugs AFAICT
17:27:19 <jaypipes> Ravikumar_hp: rock on, thx
17:27:37 <jaypipes> dwalleck: I should have the authorization tests merge prop reviewed within an hour.
17:27:49 <dwalleck> awesome, thanks
17:28:29 <jaypipes> Ravikumar_hp, donaldngo_hp: care to review the security groups merge proposals from rajalakshmi and sapan?
17:28:56 <donaldngo_hp> sure
17:29:20 <jaypipes> all: the top merge proposal is from Eoghan Glynn. It adds retries to the rest client to deal with ratelimit middleware, if it is enabled in the environment
17:29:53 <jaypipes> I've already reviewd it. looks simple enough and potentially very useful (though personally, I destroy ratelimit middlware ASAP on my envs ;)
17:30:09 <dwalleck> I just saw that. His patch should work, but the one he'll hit first is posts, which is x per hour...
17:30:36 <dwalleck> Not sure if people want their tests to pause for an hour. And the 50 per day would be even worse
17:30:43 <jaypipes> :) indeed...
17:30:54 <jaypipes> anyway, please add your comments on that review.
17:30:59 <dwalleck> But it does seem to work. Worth a try
17:31:04 <jaypipes> which leads to the last one ... https://review.openstack.org/#change,4739
17:31:15 <jaypipes> whcih is dwalleck's improvements tot he config in tempest
17:31:46 <jaypipes> it is a large change and will likely cause a lot of merge conflicts for any branch unlucky enough to try merging after it :)
17:31:53 <dwalleck> Yes...this is my start to getting the configurations organized in a more logical manner
17:32:12 <jaypipes> so... dwalleck, if you don't mind, I'll put the merge hell in your court and attempt to clear that patch last?
17:32:33 <dwalleck> jaypipes: Fair enough
17:32:55 <jaypipes> dwalleck: but by the end of the week. that means if folks don't get their reviews and review comments fixed, they'll be dealing wth those conflicts themselves :)
17:33:32 <jaypipes> OK, let's switch topics to Swift, eh? :)
17:33:38 <dwalleck> also fair :) I realize its a huge change, but I saw very clearly this week the pains people are having
17:33:45 <jaypipes> #topic JoseSwiftQA and dwalleck to give status report on swift tests
17:34:13 <dwalleck> JoseSwiftQA: We're close, right? Both for swift and CBS?
17:34:33 <JoseSwiftQA> correct.
17:34:41 <jaypipes> dwalleck: let's focus on swift right now :)
17:34:48 <dwalleck> Fair enough
17:34:53 <jaypipes> JoseSwiftQA: you guys have been adding to tempest, right?
17:35:18 <JoseSwiftQA> Service is figured out, just have to work out a few kinks and clean it up.  Haven't commited anything yet.
17:35:33 <jaypipes> k... looking forward to it!
17:35:36 <JoseSwiftQA> :)
17:35:38 <dwalleck> They weren't doing it in Tempest at first, but they're merging it all in
17:35:52 <dwalleck> I've seen it, it'l be a huge help
17:36:03 <jaypipes> Ravikumar_hp, donaldngo_hp: what about Swift at HP? any tests been worked on that could be aligned with tempest?
17:36:37 <Ravikumar_hp> we have tests , but we need to refactor little bit
17:36:46 <Ravikumar_hp> to suit temptest
17:36:55 <jaypipes> Ravikumar_hp: actually, better question might be who is the lead QA person at HP for Swift stuff?
17:37:12 <jaypipes> I can reach out to them to coordinate/collaborate with Jose and Daryl's teams
17:37:18 <Ravikumar_hp> John Lenihan in Ireland
17:37:34 <jaypipes> gotcha. OK, I'll reach out... let him know what's happening in the communtiy
17:37:46 <Ravikumar_hp> ok
17:37:54 <jaypipes> dwalleck, JoseSwiftQA: any ETA on swift stuff?
17:38:37 <jaypipes> they are hedging their bets :)
17:38:44 <jaypipes> ok, I won't push.
17:38:48 <dwalleck> JoseSwiftQA: how do you feel? I don't want you to rush
17:39:04 <dwalleck> I'm all about stability this week :)
17:39:15 <JoseSwiftQA> it's almost ready
17:39:22 <dwalleck> We don't have to have everything done, just base service and a few tests even
17:39:32 <JoseSwiftQA> Just need to find time to finish baking it
17:39:37 <JoseSwiftQA> :D
17:39:40 <jaypipes> no problem guys
17:39:46 <dwalleck> Once we get that in, everything else can follow, plus others can join in
17:40:06 <jaypipes> in the meantime, I'll get with jeblair and mtaylor about the needs for tempest in the CI infrastructure
17:40:48 <jaypipes> #topic Open Discussion
17:41:08 <jaypipes> Anybody got stuff to bring up? Any concerns? Questions? Comments about the design summit coming up?
17:41:25 <jaypipes> Does everyone have a registration code for the design summit that needs one?
17:41:33 <dwalleck> I think most of my team will be at the summit, which should be fun :)
17:41:35 <jaypipes> If not, please email me.
17:41:45 <jaypipes> dwalleck: awesome!
17:41:58 <jaypipes> donaldngo_hp and Ravikumar_hp are pretty close, so I hope they make it there :)
17:42:05 <Ravikumar_hp> we have registered . (Ravi, Nayna , Donald)
17:42:10 <jaypipes> excellent
17:42:34 <Ravikumar_hp> jaypipes: You will be in SF this week?
17:42:36 <JoseSwiftQA> We've mostly all registered.
17:43:00 <dwalleck> I also pinged Thierry about having at least one Tempest specific session, which sounds possible
17:43:01 <jaypipes> Ravikumar_hp: yep, next Tuesday in SFO and then Wed through Friday in Santa Clara at PyCon
17:43:21 <Ravikumar_hp> may be you should stop by in Cupertino
17:43:37 <jaypipes> dwalleck: we will have >1 session I believe. at a bare minimum I think we should have a Tempest Install-n-Run fest!
17:43:45 <jaypipes> Ravikumar_hp: I'd love to! :)
17:44:18 <jaypipes> OK, folks, any other questions/concerns? I'll wrap things up if not...
17:44:43 <dwalleck> I'm hoping for even more than one. I'd really like to roadmap out the future
17:44:49 <dwalleck> nope, I'm good
17:44:57 <jaypipes> ++
17:45:10 <jaypipes> OK all, have a great day. See you on the reviews :)
17:45:14 <jaypipes> #endmeeting