Friday, 2017-03-10

*** svenkat has quit IRC00:06
*** tblakes has joined #openstack-powervm00:18
*** smatzek has joined #openstack-powervm00:18
*** smatzek has quit IRC00:19
*** smatzek has joined #openstack-powervm00:19
*** smatzek has quit IRC00:24
*** smatzek has joined #openstack-powervm00:27
*** dwayne has quit IRC00:45
*** k0da has quit IRC00:46
openstackgerritgengchc2 proposed openstack/ceilometer-powervm master: Remove support for py34 for ceilometer-powervm
*** thorst has joined #openstack-powervm00:53
*** svenkat has joined #openstack-powervm01:03
*** tblakes has quit IRC01:35
*** thorst has quit IRC01:45
*** smatzek has quit IRC02:12
*** seroyer has joined #openstack-powervm02:40
*** esberglu has joined #openstack-powervm03:10
*** thorst has joined #openstack-powervm03:13
*** esberglu has quit IRC03:15
*** thorst has quit IRC03:18
*** svenkat has quit IRC03:31
*** seroyer has quit IRC03:42
*** esberglu has joined #openstack-powervm04:10
*** thorst has joined #openstack-powervm04:14
*** esberglu has quit IRC04:15
*** thorst has quit IRC04:18
*** thorst has joined #openstack-powervm05:15
*** thorst has quit IRC05:19
*** esberglu has joined #openstack-powervm05:55
*** esberglu has quit IRC05:59
*** k0da has joined #openstack-powervm06:10
*** k0da has quit IRC06:47
*** esberglu has joined #openstack-powervm06:55
*** esberglu has quit IRC07:00
*** thorst has joined #openstack-powervm07:16
*** thorst has quit IRC07:21
*** edmondsw has joined #openstack-powervm07:39
*** edmondsw has quit IRC07:44
*** esberglu has joined #openstack-powervm07:50
*** esberglu has quit IRC07:54
*** thorst has joined #openstack-powervm08:17
*** thorst has quit IRC08:22
*** k0da has joined #openstack-powervm08:23
*** alainfisher has joined #openstack-powervm09:11
*** thorst has joined #openstack-powervm09:18
*** thorst has quit IRC09:23
*** esberglu has joined #openstack-powervm09:35
*** esberglu has quit IRC09:39
*** edmondsw has joined #openstack-powervm10:04
*** edmondsw has quit IRC10:09
*** thorst has joined #openstack-powervm10:19
*** thorst has quit IRC10:23
*** openstackgerrit has quit IRC10:33
*** esberglu has joined #openstack-powervm10:34
*** esberglu has quit IRC10:39
*** thorst has joined #openstack-powervm11:20
*** thorst has quit IRC11:24
*** esberglu has joined #openstack-powervm11:30
*** esberglu has quit IRC11:38
*** smatzek has joined #openstack-powervm11:58
*** jpasqualetto has joined #openstack-powervm12:23
*** k0da has quit IRC12:28
*** seroyer has joined #openstack-powervm12:40
*** k0da has joined #openstack-powervm12:41
*** thorst has joined #openstack-powervm12:43
*** seroyer has quit IRC13:03
*** esberglu has joined #openstack-powervm13:12
*** esberglu has quit IRC13:13
*** esberglu has joined #openstack-powervm13:13
*** esberglu has quit IRC13:17
*** edmondsw has joined #openstack-powervm13:23
*** seroyer has joined #openstack-powervm13:43
*** k0da has quit IRC13:47
*** tblakes has joined #openstack-powervm13:50
*** seroyer has quit IRC13:57
*** apearson has joined #openstack-powervm13:59
*** seroyer has joined #openstack-powervm13:59
*** k0da has joined #openstack-powervm14:04
*** alainfisher_ has joined #openstack-powervm14:16
*** alainfisher has quit IRC14:18
*** alainfisher_ is now known as alainfisher14:18
*** svenkat has joined #openstack-powervm14:21
*** tjakobs has joined #openstack-powervm14:25
*** esberglu has joined #openstack-powervm14:27
*** smatzek has quit IRC14:29
*** esberglu_ has joined #openstack-powervm14:39
*** esberglu has quit IRC14:42
*** smatzek has joined #openstack-powervm14:52
efriedthorst esberglu_ can we discuss ?15:31
thorstefried: wait a tick15:31
thorstwe always call close_vterm...even if there wasn't one open15:31
thorstand it never error'd out before15:32
thorstI think this is a scenario where we're somehow calling close_vterm after the VM is deleted.15:32
efriedOr before it ever existed, yes.15:32
esberglu_I think it never existed15:32
thorstso why don't we just do a quick 'partition exists' check?15:32
efriedBut... that would _probably_ work the vast majority of the time.15:33
thorstI mean its a dlt_lpar15:33
efriedNo, I think we have a more consistent solution.15:33
thorstso I don't think its a timing thing15:33
efriedDiscussion esberglu_ and I had yesterday after you left.15:33
efriedIf you look at the other methods in - e.g. power_on/_off - they've got 404 trappers in them that no-op.15:34
efriedSo we felt it would be appropriate if dlt_lpar did the same thing.15:34
thorstright, but you'll just fail at line 65215:34
efriedAnd then we can remove this horrible horribleness from the driver:
thorstnot all of it...15:35
thorstmaybe to 69615:35
esberglu_Yeah just to 69615:35
thorstplus, I think that the dlt_lpar is the last thing to get called there15:35
thorstwhere as that 'doesn't exist' exception would probably get raised earlier in the task flow15:35
efriedCause power-off already no-ops it.15:36
efriedAnd delete & disconnect disks will still work.15:36
efriedI'll look, but likewise no-ops already in place.15:36
efriedElse we wouldn't have seen this error, would we?15:36
thorstI guess I like the idea of checking up front (because task states should sync ops...but I know that's not always valid)15:37
efriedEspecially if the LPAR never existed, no VIFs would have been created, so UnplugVifs would no-op for that reason15:37
thorstso having this crappy except block kinda caught it in case any op failed.15:37
efriedthorst Only for dlt_lpar.  Which was fine if it was the actual DELETE that raised 404, because it was using the URI of that format.15:38
efriedAnd in theory, that trap should catch the 404 from the CloseVTerm - but it's not, because the URI format isn't the same.15:38
thorstefried: unless vifs (which have CNA's that hang under)....but I see your point15:38
efriedThis comparing on the URI format is ugly, stupid, gorpy, and wrong.15:38
thorsto totes...agree with that15:38
thorstI hated that logic when it went in15:38
efriedAnd the 404 check should be moved down into the vm lib method.15:38
thorstI just have to argue otherwise now15:38
efriedbecause that matches the paradigm of the other methods in that lib.15:39
thorstline 65215:39
thorstwouldn't that fail?15:39
efriedWith a 404.  If we got to it.  Yes.15:39
efriedBut I contend that, if the CloseVTerm 404s, we just quit right then.15:39
thorstheh...just to be a jackass15:40
efriedBoth of those ops should be wrapped in 404 handlers.15:40
thorstor the same 404 handler...15:40
efriedUh, I'm good with the timing there.15:40
efriedThe timing problem I was mentioning before was, if we check and the LPAR exists, and then it gets deleted between then and when we call dlt_lpar, we'll get the 404 anyway.15:40
efriedBut here, if we get a 404, the LPAR is already gone.15:41
efriedThe likelihood that it's still getting created somewhere...  I don't think we have that code path, do we?15:41
thorstprobably not15:41
efriedAnyway, if you prefer, we ignore the 404 from CloseVTerm, and then try the delete anyway, and ignore its 404 as well.15:41
efriedThat would certainly be... thorough.15:41
thorstno no...I'm just saying, widen the scope of the try catch15:41
thorsthave the close and dlt in the same try block15:42
efriedThat will be tricky, though.15:42
thorstthen if either 404...well the VM is deleted15:42
thorstand you're good15:42
efriedBecause if we catch an exception that's _not_ 404 - like when there's no VTerm in the first place - we would still want to proceed to the delete.15:42
efriedSo I think we need two try/except blocks, unfortunately.15:42
thorstbut what I'm saying is, close_vterm doesn't error when there is no vterm in the first place15:43
thorstotherwise 99% of all deploys always would have failed.15:43
efriedOkay.  But what if it errors for some other reason?15:43
efriedSmall chances, I know.  Just trying to be safe.15:43
thorst:-)  we could wrap most code in try/excepts that way15:43
thorstI think that's paranoia15:43
efriedDoesn't mean they're not out to get me.15:44
thorstthen in 6 months you'll be saying 'and we can get rid of this ugly except block'15:44
* thorst too far? nah...15:44
efriedShall I swear by the moon?15:44
thorstuhh?  reference lost on em15:44
efriedTh'inconstant moon / That monthly changes in her circle orb15:45
efriedRomeo & Juliet.  Educate yourself <slap>15:45
*** seroyer has quit IRC15:45
efriedWe scared off seroyer15:45
thorstwell now you made seroyer leave15:45
thorstanywhoooo, I think its a little paranoid to do that try/except.  I think its fair to do it all in one.15:46
thorstI'll put a comment in15:46
efriedOkay.  So esberglu_ you up with the plan?15:46
efriedRemove the horrible hack from the driver; put a try/except to catch and ignore 404 into dlt_lpar.15:47
esberglu_Rip out that crappy URI exception thing and have 1 404 exception for vterm close and delete15:47
efriedthorst Is there any actual reason for us to be testing the URI?15:48
efriedIf we get a 404, that's definitive, neh?15:48
thorsttesting the URI?15:48
efriedString matching horribleness.15:48
thorstuhh, didn't you make that/15:48
efriedNot a chance.15:48
efriedLet me look.15:48
* thorst scurry to gitblame15:49
efriedI'll just leave that up there, see if he deigns to answer for himself.15:50
kylek3hon, but presenting at the moment.  will check later.15:50
efriedI'll ask apearson if there's any reason for us to be doing that.15:50
thorstyou probably doesn't hurt to have a lpar_exists check just at the beginning of destroy itself.  Half second (if that) check...15:50
thorstbut meh15:50
thorstefried: thoughts?15:51
efriedMeh indeed.15:51
efriedProbably not worth the effort of thinking through all the implications for LPM, RR, etc.15:51
thorsthow would it impact any of those?15:52
thorstI suppose it could impact the disk15:52
efriedPause, suspend, shelve...15:53
thorstfine fine15:53
thorstpoint taken15:53
efriedOr Rebuild - doesn't that deliberately delete the disks without deleting the LPAR, or vice versa, or something?15:54
efriedYeah, anyway, that would be a different change set, methinks.15:54
efriedWrite up a blueprint for it.15:54
thorstfair enough15:54
*** alainfisher has left #openstack-powervm16:00
*** seroyer has joined #openstack-powervm16:01
efriedesberglu_ What was that change set you said needed to be merged before our CI would start working again?16:07
esberglu_There's actually 2 we need, 1 for tempest, 1 for g-r ocata16:07
efriedoh, that was just ocata, right.16:09
efriedthorst (was looking for change sets with no reviewers).  See any reason not to accept this?16:19
esberglu_Also the same g-r change is needs to merge for newton CI16:20
adreznecefried: thorst are we going to merge ? Or are we just going to wait on g-r until we figure out pypowervm verisoning16:22
efriedadreznec Wait.16:22
adreznecMy vote personally is just to merge it, but...16:22
efriedOr merge.16:22
efriedAs you wish.16:22
thorstefried: LGTM16:23
thorstI +2'd16:23
efriedk, merging.16:23
efriedadreznec I'm okay to merge it, but that's going to allow the others in, which means we'll start getting new proposals from the bot.16:23
adreznecOne thing I did notice while looking at that now is that it's removing the discover req from test-requirements16:24
efriedI buzzed jfoliva in slack earlier, but haven't heard back.16:24
adreznecWe passed CI without it... any idea what that req was for offhand?16:24
adreznecefried: I think we're going to start getting bot proposals regardless16:24
adreznecNow that has a WF+116:25
efriedThat guy has a depends-on to the above, so it won't merge yet.16:25
adreznecAh, I see16:25
adreznecMissed that16:25
*** thorst is now known as thorst_afk16:27
*** k0da has quit IRC16:30
*** openstackgerrit has joined #openstack-powervm16:35
openstackgerritMerged openstack/nova-powervm master: Use ConfigParser instead of SafeConfigParser in Python 3
*** seroyer has quit IRC16:44
efriedthorst_afk ??16:48
openstackgerritEric Berglund proposed openstack/nova-powervm master: Except HttpError 404 dlt_lpar
*** nbante has joined #openstack-powervm17:05
efriedesberglu_ IOError: [Errno 2] No such file or directory: '/opt/stack/tempest/.tox/tempest/lib/python2.7/site-packages/appdirs-1.4.0.dist-info/METADATA' Is this because of that setuptools BS?17:07
efriedbutbutbut... I have 33.1.1 on my system.17:08
efriedAnd I'm not on ocata, dammit.17:09
efriedguess it doesn't matter.17:09
*** seroyer has joined #openstack-powervm17:12
efried'cept the corresponding change set in master has already been merged.  So that must not be my problem?17:14
efriedPerhaps I need to upgrade to 34-something-that-isn't-forbidden?17:15
*** nbante has quit IRC17:50
esberglu_efried: Sorry I totally missed this earlier. The corresponding tempest change hasn't merged yet19:19
esberglu_So you have the g-r change but not the tempest change19:20
esberglu_efried: adreznec: thorst_afk: It was brought to my attention that when we leave 'recheck powervm' comments, we are actually kicking off rechecks for ALL CI systems19:21
esberglu_Because they just check for 'recheck *'19:22
esberglu_I have a proposed change up (4974)19:22
esberglu_That updates or recheck trigger to 'powervm recheck' (among a few other variants)19:22
esberglu_And will also trigger on 'recheck *'19:23
esberglu_To bring us in line with other CIs19:23
thorst_afkesberglu_: I'm good with the change, just be sure to also update the OpenStack CI page with the recheck comment19:23
esberglu_I updated the comment that CI posts to include the updated recheck trigger. Is that what you mean? Not sure where else we have the recheck comment format posted19:25
esberglu_Unless you are talking about our PowerVM CI wiki page. (which doesn't have the old format)19:28
thorst_afkyeah, that was it19:29
*** esberglu_ has quit IRC19:39
*** esberglu has joined #openstack-powervm19:39
*** esberglu has quit IRC19:44
*** esberglu has joined #openstack-powervm19:51
*** k0da has joined #openstack-powervm20:22
efriedesberglu I would actually rather not trigger on 'recheck *'20:46
efriedCause I want to be able to recheck e.g. Jenkins without spinning off a whole PowerVM CI run.20:46
*** k0da has quit IRC20:51
esbergluefried: Was talking this morning about that with the PowerKVM CI operator. From that I thought it was supposed to be supported by the community21:17
esbergluAfter further discussion that is not the case21:17
esbergluI will have a change up shortly21:17
efriedthorst_afk adreznec Here's a cool thing you may not have known.21:18
efriedtox -e pep8 -- -HEAD21:18
efriedThe -HEAD bit will restrict analysis to files you've changed.21:18
efriedUseful for big projects like nova.21:18
thorst_afkthat is neat21:18
esbergluefried: 497721:19
adreznecNice, that's handy21:20
efriedesberglu I posted some comments on the already-merged one you did earlier21:21
efriedabout your regex.21:21
efriedI'll repost on this 'un.21:21
efriedesberglu Also, I don't see how what you've done here is going to help.  Maybe I'm misunderstanding the purpose of this change set...21:22
efriedoh, I get it.  Yeah, I don't like this either.21:23
efriedadreznec Help me out here.21:23
efriedI'm of the opinion that we shouldn't recheck PowerVM CI unless it's explicitly asked for.21:23
efriedI want to be able to post 'recheck' and have it just redo jenkins.21:24
efriedIs there some other comment format that will run just jenkins, but not trigger PowerVM CI?21:24
adreznecefried: Not that I'm aware of, and at least in Nova most other CIs seem to also recheck on a generic recheck21:25
efriedadreznec Well, I don't like it.21:25
efriedWe should buck the system.21:26
*** thorst_afk has quit IRC21:26
efriedDoes our CI post that handy-dandy comment explaining how to trigger a recheck, when it fails?21:26
efriedesberglu ^21:26
efriedJust knowing how relatively small our CI env is, especially since we're non-voting, I wouldn't want every recheck in the world to trigger us.21:27
efriedI think we should only go when 'powervm' is included in the comment.21:27
adreznecefried: It does21:27
efriedadreznec I'm saying we should *only* go when 'powervm' is included in the comment.21:28
efriedOh, you were answering the question about the explanatory comment when the CI fails.21:28
*** k0da has joined #openstack-powervm21:29
efriedSo yeah - convince me that it should be otherwise.21:29
adreznecI don't have a strong preference either way21:30
adreznecJust saying that's how it is with existing other CIs21:30
adreznecThe only downside is usability21:30
adreznecYou can't do a general "recheck" and hit everything. You'd have to go find the magic incantation for each21:31
efriedadreznec I'm good with that, at least for right now, because a) we're non-voting, and b) we're the only ones who care about our CI.21:31
efriedSo better to reduce the load.21:31
esbergluefried: It's one of the requirements of 3rd party CI that you support the 'recheck' comment triggering your CI21:32
efriedSo just tweak the regex to make me happy and we'll call it good.21:32
esbergluYep already did21:33
esbergluOh nvm I see what you mean21:33
*** smatzek has quit IRC21:36
*** thorst_afk has joined #openstack-powervm21:44
efriedesberglu If we merge this, how long before it takes effect?21:50
efriedCause I got one I want to recheck powervm, but don't want it to retrigger the world.21:50
esbergluWell you can already do that with 'powervm: recheck'21:51
esbergluBut I can update it in like a second after it merges21:51
efriedSo wait, what's changing here?21:51
efriedJust that before if I said 'recheck foo' it would trigger our CI?21:52
efriedAnd now even 'recheck powervm' won't do it.21:52
efriedSo it has to be 'powervm recheck'21:52
efriedUhm.  Did you update the text of that comment accordingly?21:52
efriedor did it already say the right thing.21:52
esbergluYeah I did in the 1st patch21:53
efriedokay, I see it.21:53
efriedesberglu Okay, merge away.21:54
esbergluCool, its live21:55
*** svenkat has quit IRC22:01
openstackgerritEric Berglund proposed openstack/nova-powervm master: Except HttpError 404 dlt_lpar
*** tjakobs has quit IRC22:16
efriedthorst_afk Are you out for the count at this point?22:31
thorst_afkefried: yes22:38
* adreznec waves goodbye to thorst_afk 22:42
*** seroyer has quit IRC22:43
efriedesberglu I'd like you to port in-tree once the CI verifies it.22:44
efriedesberglu I believe you'll need to stack it on top of
efriedLet me know if you need a hand with the git paperwork on that.22:45
efriedAlmost wonder if you should just submit it as a patch to that change set instead...22:46
esbergluYeah it's not that big and isn't out of place there22:56
*** tblakes has quit IRC23:12
*** apearson has quit IRC23:20
*** thorst_afk has quit IRC23:27
*** edmondsw has quit IRC23:35
*** edmondsw has joined #openstack-powervm23:37
*** edmondsw has quit IRC23:42
*** nbante has joined #openstack-powervm23:47
*** nbante has quit IRC23:51
*** tblakes has joined #openstack-powervm23:57
*** jpasqualetto has quit IRC23:58

Generated by 2.14.0 by Marius Gedminas - find it at!