16:01:49 <scottda> #startmeeting cinder_testing 16:01:50 <openstack> Meeting started Tue Jan 17 16:01:49 2017 UTC and is due to finish in 60 minutes. The chair is scottda. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:51 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:54 <openstack> The meeting name has been set to 'cinder_testing' 16:02:04 <erlon> hey 16:02:10 <scottda> ping eharney, xyang1, gouthamr, akerr, smcginnis, cFouts, e0ne, geguileo, dulek, flip214, ntpttr patrickeast, _alastor_, DuncanT, erlon, josemello,mdarnell, karthikp 16:02:22 <e0ne> hi 16:02:31 <scottda> hi erlon . Now that you are hear, we know we'll have something to talk about :) 16:02:34 <xyang> hi 16:02:48 <scottda> #link https://etherpad.openstack.org/p/cinder-ocata-testing 16:02:58 <scottda> I pointed this out last week ^^^ 16:03:03 <erlon> scottda: haha, 16:03:20 <scottda> Some tracking so that we can make sure things get *some* coverage before the release 16:03:35 <erlon> scottda: nothing much new, Im returning this week from my pto 16:03:51 <scottda> But I don't see much action on that...I'm going to put that link on the main Cinder meeting agenda 16:04:15 <erlon> scottda: I notice that etherpad is very empty, may be folks should update it with details 16:04:29 <scottda> yup 16:05:25 <erlon> scottda: I added some bullets but still need to track the migration related bugs and try to sort then out 16:06:43 <scottda> OK, I added a reminder in the main Cinder agenda 16:07:04 <erlon> scottda: nice 16:07:36 <scottda> I also have added something to the PTG etherpad regarding a single Testing priority for Pike release... 16:07:52 <erlon> scottda: people should really give some attention to testing. The AA patches are landing and they touch a lot of code 16:08:13 <hemna> what's up 16:08:28 <scottda> I think if we don't have a single focus, we'll end up with the usual list of everything wrong we'd like to fix. 16:08:56 <scottda> hemna: Hi. 16:09:15 <scottda> Anyone have anything about tests they'd like to talk about? 16:09:29 <erlon> scottda: whast do you mean by single focus? 16:10:08 <scottda> erlon: Well, if we come up with the 10 areas we need to write tempest/functional tests for, none of them will get done.... 16:10:22 <scottda> erlon: If we pick one, there is hope that we can focus and perhaps make progress. 16:10:30 <scottda> Then, we could add more if there is time. 16:10:32 <geguileo> erlon: I have a fix for the manage issue 16:10:37 <scottda> smcginnis: had suggested this, and I agree. 16:10:46 <geguileo> erlon: I'm going to look into the retype with migration now 16:11:06 <erlon> scottda: hmm, I see 16:11:09 <scottda> But, of course, people can work on whatever they want (or whatever their employer lets them do) 16:11:22 <openstackgerrit> Merged openstack/os-brick: Fix a wrong indentation https://review.openstack.org/421137 16:11:23 <erlon> geguileo: I remember I have some patches but didnt submit 16:11:43 <erlon> geguileo: ill put a patch and you can give a look if theres something missing 16:11:51 <scottda> geguileo: I had left some work undone on retype with migration...we hope we can end up calling Nova swap volume, to test that as well. 16:11:57 <geguileo> erlon: lol, then our patches will be fighting one another 16:12:28 <erlon> geguileo: haha 16:12:48 <scottda> geguileo: erlon I think that using LVM will cause a Nova swap_volume if the hostname for both backends is different. Otherwise, it won't call swap_volume. 16:13:52 <geguileo> scottda: I have just fixed the manage one, I have to look at the other case, as it may be related 16:14:30 <erlon> scottda: I dont reacall now, but I believe swap_volume is always being called for the attached migration 16:14:40 <scottda> OK, cool. I think there is a local nova test for swap_volume, but nothing that exercises it from Cinder... 16:15:07 <scottda> and therefore not testing of anything that has Cinder call Nova, and using the novaclient, etc. 16:15:11 <markus_z> scottda: Would you maybe have a look at https://review.openstack.org/#/c/413684/ ? Needs a 2nd +2. 16:15:27 <scottda> erlon: No, I don't think it will with a single-node LVM setup, since the hostname is the smae for both backends. 16:15:31 <scottda> s/smae/same 16:16:29 <scottda> erlon: I thought I had a patch that worked, but there is some logic to optimize things for the case of 2 backends on the same host...I'll try to find it... 16:16:39 <scottda> markus_z: I'll put it in my queue to have a look 16:16:55 <markus_z> scottda: Thanks, that's all I can ask for :) 16:18:55 <openstackgerrit> Erlon R. Cruz proposed openstack/cinder: Fix sched/manager calls to filters https://review.openstack.org/421348 16:19:19 <erlon> geguileo: ^ first :) 16:19:35 <scottda> #link https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L2293 16:19:39 <geguileo> erlon: lol, I still have to add the tests 16:19:49 <geguileo> erlon: And see if it helps with the other bug 16:19:54 <scottda> erlon: That is it, I think 16:21:03 <erlon> scottda: yep, theres this case, I found it because I was testing migration between pools in the same backend 16:21:30 <scottda> erlon: So, does your test patch get around this? If not, we'll go with geguileo 16:21:31 <scottda> :) 16:21:32 <erlon> scottda: which is the must be the same behavior you mentioned in LVM 16:21:40 <erlon> scottda: yes 16:22:08 <erlon> scottda: ow wait 16:22:58 <scottda> erlon: I think we really need to check the Nova logs to verify. Otherwise, we'll need an additional test. And this already takes a bit of time to run. 16:23:10 <geguileo> erlon: your patch is incomplete 16:23:11 <openstackgerrit> Erlon R. Cruz proposed openstack/cinder: Fix retype with migration false positive success https://review.openstack.org/421353 16:23:32 <erlon> geguileo: yeap, that another error 16:23:41 <erlon> I split in those 2 16:23:52 <scottda> erlon: You don't get to be first to post a patch if it's not complete :) 16:24:08 <erlon> geguileo: scottda: the last one is handling the hostname 16:24:18 <erlon> scottda: haha damn 16:25:31 <erlon> scottda: geguileo: they where hanging here, fell free to -2 if you have another that fix 16:26:06 <geguileo> erlon: I'll look now at reproducing the second issue and do the RCA 16:27:49 <erlon> geguileo: ok, leme know if you have problem to reproduce it 16:27:56 <openstackgerrit> int32bit proposed openstack/python-cinderclient: Add snapshot-id filter to get volumes via cli https://review.openstack.org/416510 16:27:57 <geguileo> erlon: why do we need to say that None is not equal to None for host equivalency? 16:29:52 <erlon> geguileo: I don't think we need to. Actually they should not be considered equal 16:30:35 <erlon> geguileo: if clustering is not configured, both are None, but the hosts are actually different 16:30:36 <geguileo> erlon: Why should they not be considered equal? 16:31:12 <geguileo> erlon: but that should be ok 16:31:22 <geguileo> erlon: clusters are equal but not the hosts 16:31:37 <erlon> the false positive happens here:https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L2295 16:33:00 <erlon> geguileo: yeah but if does not have cluster_name set, it will lead to a false positive as will compair None == None 16:33:39 <geguileo> erlon: I'll check it, because then the problem is somewhere else, not in that method 16:35:05 <erlon> geguileo: I was trying to understand if the cluster_name that should always have the right values or if the comparation should consider that they could be None, but didn't get to a conclusion, that's why I pinged you the last time 16:37:37 <scottda> Ok, well you two can battle it out for which patch we go with.... 16:37:51 <scottda> But please check the Nova swap_volume call. 16:38:11 <scottda> Anything else about tests today? 16:39:22 <erlon> scottda: hmm, I found some tests in tempest that worth to look on 16:39:37 <erlon> scottda: https://review.openstack.org/#/q/project:openstack/tempest+(message:volume+OR+message:cinder)+status:open+-status:merged 16:40:06 <erlon> scottda: it seems a lot of people now and in the past have worked in some testing and just left the work 16:40:30 <erlon> scottda: so, some of the might be interesting to re-pick 16:40:41 <scottda> erlon: Yes. Thanks for cleaning all of those up. 16:40:42 <scottda> :) 16:41:24 <erlon> scottda: haha, welcome, in advance :) 16:41:35 <erlon> scottda: ill have a look 16:41:59 <scottda> I'm kidding, of course. But let me know if you are re-activating work on any, and I'll review. 16:42:53 * scottda notes his patch is at the top of the list 16:43:00 <erlon> scottda: ill have a look in the list and ping you about the best ones we and do what we can afford to 16:43:13 <erlon> scottda: welcome again :) 16:43:52 <openstackgerrit> int32bit proposed openstack/python-cinderclient: Add snapshot-id filter to get volumes via cli https://review.openstack.org/416510 16:44:36 <erlon> scottda: another point I was wondering weather we should pick some of that and move to our intree tests 16:45:24 <scottda> erlon: That could be a good idea (moving in-tree), as that may be why some languished in the review queue. 16:46:03 <smcginnis> +1 16:46:49 <erlon> scottda: mhm, the bad side is that we loose the eyes of tempest core, but that's is hard to get anyways 16:47:23 <scottda> Yes, the "hard to get" part is what I'm talking about 16:47:39 <smcginnis> erlon: I'm sure if we merge something especially bad, we'll get their attention. ;) 16:47:40 <erlon> scottda: yeap 16:47:54 <smcginnis> scottda: Right, it's lack of attention I think that has prevented us from adding some of these tests. 16:47:59 <erlon> smcginnis: haha, yeah 16:47:59 <scottda> Or it they merge something especially bad... 16:48:07 <smcginnis> scottda: ;) 16:48:36 <scottda> OK, well thanks erlon . Keep us informed as to what you identify (if still in Tempest). 16:48:51 <scottda> I reckon we'll pay attention no matter what if you move them to cinder in-tree 16:49:50 <scottda> OK, anything else? 16:50:16 <xyang> so regarding tempest tests for groups, should I submit them in-tree or in tempest? 16:51:10 <scottda> xyang: Good question. 16:51:12 <smcginnis> xyang: Hmm, maybe try in tempest, and if there is resistance there move to in-tree? 16:51:18 <xyang> sure 16:51:23 <scottda> +1 16:51:43 <erlon> scottda: sure, welcome 16:52:05 <smcginnis> We've talked about using our in-tree as kind of an incubator for tempest tests, but I think that one might be good to just go there. 16:52:19 <scottda> yup 16:52:43 <xyang> @smcginnis: this one works for LVM, so should be okay I hope 16:52:49 <erlon> xyang: if you are getting good feedback, and they are ready to get merged Id recommend to leave there, otherewise move then to in-tree 16:53:03 <xyang> erlon: ok 16:55:02 <scottda> OK, if there's anything else, chime in. Otherwise, thanks everyone 16:55:17 <smcginnis> Nothing here. Thanks scottda. 16:55:44 <scottda> #endmeeting