16:00:00 #startmeeting cinder 16:00:00 Meeting started Wed Mar 4 16:00:00 2015 UTC and is due to finish in 60 minutes. The chair is thingee. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:01 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:04 The meeting name has been set to 'cinder' 16:00:10 DuncanT: yes, it's now time 16:00:17 hi everyone 16:00:20 Hi! 16:00:21 o/ 16:00:21 hi 16:00:22 hey 16:00:22 hi 16:00:23 hello 16:00:25 hi 16:00:25 hi 16:00:25 Hi 16:00:26 hello 16:00:27 hi 16:00:30 hi 16:00:30 o/ 16:00:33 hi 16:00:35 рш 16:00:37 Third party ci reminder 16:00:38 hi! 16:00:39 #link http://lists.openstack.org/pipermail/openstack-dev/2015-January/054614.html 16:00:42 morning 16:00:43 March 19th 16:01:16 you should have a third party running way earlier than that. Expecting to have one done on March 19th is just not logical. 16:01:39 hi 16:01:40 Please see me if you're driver maintainer and have no idea what I'm talking about 16:01:50 Cinder deadlines... 16:01:59 #link http://lists.openstack.org/pipermail/openstack-dev/2015-February/056964.html 16:02:04 o/ 16:02:07 march 10th, all blueprint code must be merged. 16:02:14 if it's not merged by then, it's not going in. 16:02:17 no exceptions. 16:02:24 bug fixes from there on out for Kilo 16:02:57 I'll be hitting the reviews hard this week, and I hope others will join me 16:02:59 #link https://launchpad.net/cinder/+milestone/kilo-3 16:03:18 great job to all cinder contributors for this 16:03:25 we're getting close! 16:03:28 thingee, +1 16:03:33 see the priorities here https://etherpad.openstack.org/p/cinder-k3-priorities 16:03:36 of what to review 16:03:42 hi 16:03:48 even if you're not core, your review matters! 16:03:50 thingee: Did you see the discussion about the oslo logging namespace change? Ivan is going to update it on the 10th, we want to merge it last so it doesn't cause rebase conflicts for everybody else 16:04:27 DuncanT: there's an agenda item about this. we can talk about it then 16:04:30 lets get started! 16:04:31 thingee: Actually, it's later on the agenda, ignore me for now 16:04:34 #link https://wiki.openstack.org/wiki/CinderMeetings#Next_meeting 16:04:46 #topic Vendor driver documentation changes 16:04:48 smcginnis: hi 16:04:54 thingee: Hey! 16:04:57 #link https://review.openstack.org/#/c/133372/ 16:05:04 #link https://wiki.openstack.org/wiki/Documentation/VendorDrivers 16:05:15 I just wanted a moment to point out some vendor driver documentation changes. 16:05:30 See links above for the review and a new wiki page to be aware of. 16:05:45 Basically the plan is to make vendor driver documentation smaller. 16:05:55 Mostly config info hosted by OpenStack, 16:06:10 So some vendors probably will need to work on hosting their own documentation. 16:06:20 It can then be linked to from the official manuals. 16:06:28 thanks for bringing it up smcginnis! 16:06:31 Please take a moment to go to the wiki page and update your info. 16:06:36 annegentle: No problem! 16:06:48 That's really it. Just wanted to make sure everyone was aware of that. 16:07:02 annegentle: Please correct me if I've mis-stated anything. 16:07:03 This is excellent 16:07:11 smcginnis: looks good to me 16:07:19 smcginnis: and yes, really want to raise awareness 16:07:41 smcginnis: consider updating https://wiki.openstack.org/wiki/Cinder/how-to-contribute-a-driver 16:07:53 thingee: Good point, I'll take care of that. 16:08:14 annegentle, would you prefer we have one line per driver or company on the vendor driver wiki? 16:08:23 #action smcginnis to update how to contribute cinder driver doc with info on driver doc in openstack docs 16:08:24 so much doc 16:08:31 kmartin: looking 16:08:51 annegentle: If you could also review this section at some point that would be great: https://wiki.openstack.org/wiki/Cinder/how-to-contribute-a-driver#After_Your_Driver_Is_Added 16:08:53 what's the deadline for documentation changes? 16:09:22 kmartin: I like it as one line per "configurable" driver (even if the vendor's the same) 16:10:13 rhe00: for the Configuration Reference, the deadline is same as code, https://wiki.openstack.org/wiki/Kilo_Release_Schedule 16:10:57 #link http://specs.openstack.org/openstack/docs-specs/specs/kilo/move-driver-docs.html 16:11:10 ok, thanks 16:11:17 thanks smcginnis and annegentle! 16:11:28 thankee thingee! 16:11:35 :) 16:11:41 #topic Following up modifying volume image metadata 16:11:43 davechen: hi! 16:11:45 hi 16:11:55 Here is some quick update for volume image metadata modification to follow up some comments from previous meeting. 16:11:57 spec: 16:11:59 #link https://review.openstack.org/#/c/160147/ 16:12:04 cli: 16:12:06 #link https://review.openstack.org/#/c/147077/ 16:12:15 Firstly, remove the snapshot related changes both in the SPEC (https://review.openstack.org/#/c/160147/) 16:12:21 and CLI patch (https://review.openstack.org/#/c/147077/). 16:12:28 So, only CRUD on the image metadata is proposed in the patches. 16:12:47 o/ 16:12:51 Secondly, import both role based props protection (https://review.openstack.org/#/c/160090/) and policies based props protection(https://review.openstack.org/#/c/160145/) from Glance, and have a thorough testing recently. 16:13:11 Props protection is based on the RBAC, basically, role based props protection judge user's permission by his role and compare his roles with predefined rule configuration files; 16:13:20 polices based props protection parse the rules defined in configuration files and appended the rules to the basic RBAC rules and then determine the user access permission. 16:13:37 davechen: Any plan at some point to make this a common library with glance? 16:13:47 these is the most important comment I get from previous meeting. 16:14:08 DuncanT: common library? 16:14:22 DuncanT: not yet 16:14:48 davechen: Ok, thanks. Just wondered 16:14:58 davechen: I appreciate the work on this. 16:15:00 DuncanT: just did these separately, but it is good idea to do in that way. 16:15:17 so, anything else I need address? 16:15:21 davechen: at this time we'll have to revisit things in Liberty. We're already pretty late in the Kilo cycle with the 20 targeted things in review. 16:16:02 thingee: yeah, so is it still possible to merge the basic CRUD APIs? 16:16:41 davechen: The problem is review resources. If I can keep focused on what's already on our plate, that would be great. 16:16:58 davechen: plus the number of people that would be pissed that aren't getting in because of the feb 15 feature freeze deadline 16:17:00 thingee: just check the review list, the feature is not planed in the review list even if the low priority. 16:17:15 davechen: and march 1st code pre feature freeze deadline 16:18:07 thingee: okay, hope some volunteer could help to review it. 16:18:48 davechen: thanks 16:18:55 #topic OpenStack Infra: add non-voting Sheepdog test job to cinder check pipeline 16:18:57 kazum: hi 16:19:02 hi 16:19:07 I'm not sure why we're discussing this here. Not much we can do. 16:19:24 i didn't understand hemna's objection on this one 16:19:40 my objection was related to the fact that this was added w/o talking to the Cinder team 16:19:51 they want to make it voting 16:19:59 if it is stable and non-voting, why not? 16:20:03 No way is it going to be voting 16:20:12 Their entire intention was to make it voting 16:20:12 the commit message says never voting 16:20:18 i thought it was aiming for voting, which would put it in line with what we are already doing for ceph/gluster 16:20:21 and they didn't even talk to us about it. 16:20:23 non-voting* 16:20:25 yes, we have no plan to make it voting. 16:20:35 so this isn't really a new scheme 16:21:01 So I know there was some discussion about this on the ML. I really don't care if people want to add checks in the pipe queue if infra wants to work with them. 16:21:16 Bringing it up at the meeting is probably polite, just so there's an opportunity to shout. 16:21:17 Original commit mentioned planning to eventually vote. That was my only concern. 16:21:18 non-voting for sure. 16:21:25 smcginnis, +1 16:21:35 I'd rather it didn't vote as jenkins, but it isn't the end of the world 16:21:38 yup, their original messaging was to make it voting 16:21:52 Any concern with this increasing gate time? 16:21:55 hemna: it might've just been a mistake 16:21:56 hemna: yes, sorry for confusion. 16:21:58 so I just thought it wasn't very cool to do something like this without at least talking to the Cinder team first. 16:22:04 and misunderstanding 16:22:09 that's all. 16:22:19 Voting as jenkins makes it look more official than the other 3rd party CIs, which I don't really like 16:22:31 fortunately infra is waiting for Cinder folks to give feedback on the review 16:22:45 DuncanT, I agree with you 16:22:50 eharney, we only found out about it because anteaya brought it up to me 16:22:51 Would be great if there was a separete open source third party CI. 16:22:57 well I think the interested parties should follow up on the review. Nothing should be voting. 16:23:01 We've discussed this here: 16:23:01 thankfully she gave us a heads up 16:23:03 #link https://wiki.openstack.org/wiki/Cinder/tested-3rdParty-drivers#When_thirdparty_CI_voting_will_be_required.3F 16:23:20 infra is working on creating new accounts to comment as not-jenkins 16:23:42 hemna: can you follow up so infra feels comfortable again? 16:23:48 thingee, sure 16:24:01 kazum: anything else? 16:24:05 all I needed to see was that you knew 16:24:07 just to be clear 16:24:09 so now I see 16:24:11 and thanks 16:24:13 asselin_: Cool stuff 16:24:13 we are ok with it non-voting ? 16:24:14 no, thanks a lot :) 16:24:42 hemna: right non-voting 16:24:46 ok cool. 16:24:49 current decision still stands at https://wiki.openstack.org/wiki/Cinder/tested-3rdParty-drivers#When_thirdparty_CI_voting_will_be_required.3F 16:24:53 until rediscussion 16:25:00 +2 16:25:14 hemna, kazum: thanks 16:25:35 #topic Porting volume drivers to use ABCmeta 16:25:40 mkoderer: hi 16:25:41 hi there :) 16:25:42 DuncanT: hi 16:25:54 Hi 16:26:01 #link https://review.openstack.org/#/c/138999/ 16:26:01 mkoderer has the mike 16:26:08 #link https://review.openstack.org/#/q/project:openstack/cinder+branch:master+topic:bp/abc-volume-drivers,n,z 16:26:09 the driver change already got merged 16:26:19 #link https://review.openstack.org/#/c/127455/ 16:26:19 and I just realized that the BP was in state implemented 16:26:34 but in fact there is still much to do 16:26:48 IMHO we should try to port all drivers to use the model 16:27:03 mkoderer, in K ? 16:27:04 so I started with: rbd, gluster, nfs and block_device 16:27:05 mkoderer: +1 16:27:08 mkoderer ++ 16:27:09 in liberty right? 16:27:20 mkoderer: would consider breaking up the blueprint then. Have the base driver class bp marked as implemented. 16:27:27 I personally don't care... it's up to you guys :) 16:27:30 Might as well start ASAP, it is a mostly a mechanical change 16:27:32 patrickeast, +1 16:27:34 mkoderer: just in interest of time left in kilo with testing 16:27:48 I'm a bit worried making this change this late in the game 16:27:53 hemna: +1 16:28:03 I'd vote for doing it early in L 16:28:08 so the changes as mostly simple: https://review.openstack.org/#/q/project:openstack/cinder+branch:master+topic:bp/abc-volume-drivers,n,z 16:28:21 hemna: +1 16:28:33 mkoderer: we can't verify it without CI 16:28:34 ok let's to it in L 16:28:43 changes may be simple but soak time is needed 16:28:45 e0ne: I see 16:28:51 tbarron, yup 16:28:57 mkoderer: i mean 3rd party ci :( 16:28:59 Agree this shoudl wait for L. 16:29:08 mkoderer, yes but all the additional reviews for each driver change needs to be taken into account 16:29:14 +1 for L 16:29:29 thingee: should I create an additional spec for that? 16:29:36 tbarron: +1 16:29:37 #agreed we wait for the rest abc driver changes in L 16:29:38 L 16:29:38 mkoderer: please 16:29:42 thingee: k 16:29:54 #action mkoderer to break up abc driver blueprint 16:29:57 mkoderer: Is this something you are planning to do or do driver owners need to plan to make the changes? 16:30:01 mkoderer, I'd like to ping you about a driver.FibreChannelVD in #cinder and see if it makes sense to you 16:30:07 #action to adjust kilo release page accordingly with blueprints 16:30:12 and I added a script that extracts the driver support matrix out of ABC 16:30:13 https://review.openstack.org/#/c/160346/ 16:30:37 jungleboyj: I would really like to have the help of the driver owners .. yes :) 16:30:43 jungleboyj, mkoderer: I have plans to send a mass email to driver maintainers on some other changes in Cinder, so it would be great to send it out then as well. 16:30:57 hemna: ok do that 16:31:11 mkoderer: Ok. Thanks for the heads up. 16:31:14 thingee: Good idea. 16:31:26 thingee: +1 16:31:26 mkoderer, we can convert our drivers, so no problem there. 16:31:32 just in L :) 16:31:35 thingee: Can you cc me on any such mass emails. please? I invariably get asked on IRC about them the day after (since I'm in a GMT+X timezone) and I have to go digging 16:31:41 emails about new things in Kilo that drivers can do in L will be going out soon. Want to make sure folks are informed. 16:31:43 hemna: :) ook 16:31:45 mkoderer: Same here, just need to example to work from. 16:32:04 DuncanT: sure 16:32:13 Thanks 16:32:14 I also send a copy to the mailing list though 16:32:14 jungleboyj: yep I guess there are enough examples already in.. and it should be simple 16:32:17 DuncanT: ^ 16:32:27 did te same thing with the CI email 16:32:31 jungleboyj: like https://review.openstack.org/#/c/138661/ 16:32:41 thingee: That's where I go digging for it. Anything with my address CCd bypasses filters, that's all 16:32:49 DuncanT: k 16:32:51 mkoderer: Cool. Thanks! 16:32:56 k folks thx 16:33:01 mkoderer: thank you 16:33:26 mkoderer: we all appreciate your work in this release with getting things going :) 16:33:51 #topic Move to oslo_log is coming! 16:33:55 thingee: I do my very best :D 16:33:55 jungleboyj: hi 16:33:59 thingee: tha main question is: do you want to land it in Kilo? 16:34:07 #link https://review.openstack.org/#/c/157441 16:34:18 when is the best time for it? before or after FF. it requires a lot of rebases 16:34:19 thangp: Hi. 16:34:28 thingee: Hi. 16:34:29 jungleboyj: hi 16:34:37 thangp: Sorry, wrong th* 16:34:41 lol 16:34:46 :) 16:34:47 e0ne, jungleboyj: what does it buy us to rush things? 16:35:08 thingee: So, I think we want to get this landed in K. DuncanT and hemna Were feeling that way. 16:35:28 thingee: personally, i've got a lot requests from our operators for it 16:35:29 thingee: Less that we need to worry about oslo.log in incubator. 16:35:34 thingee: It is a good time to do it, straight after code freeze - otherwise we get lots of merge conflicts in future 16:35:37 jgriffith: Had requested it. 16:35:48 DuncanT: +1 16:35:49 avoiding rebase churn would be nice. 16:36:00 I'm trying to avoid that as well with the shadow tables review 16:36:05 hemna: +1 16:36:09 e0ne: requests about what? 16:36:16 So, we would like to do it 3/10 or 3/11 to avoid disrupting the new function and get it in before we start doing a lot of fixes. 16:36:20 DuncanT: that's compelling 16:36:32 well technically this is bug in Kilo https://launchpad.net/cinder/+milestone/kilo-3 16:36:34 Still get plenty of runt ime on it. 16:36:35 winston-d: moving to oslo.log to make logs of different components similar 16:36:40 so I don't think anyone needs to convince me ;) 16:36:43 * DuncanT is happy to +2 it very early on the 11th 16:36:59 thingee: i filled a BP for it 16:37:07 Just so that it doesn't stop any last minute merges on the 10th 16:37:08 e0ne: not sure why we need one 16:37:17 thingee: +2 16:37:27 anyways, I'm fine if we can get it early after the freeze to help avoid rebase hell 16:37:36 thingee: +2 16:37:49 thingee: thanks! +1 for each rebase 16:37:57 Great. Mainly I just wanted to bring it up here so people knew it was coming and there would be a brief period of rebase hell. 16:38:03 In a week. 16:38:04 jungleboyj: makes sense 16:38:30 jungleboyj, e0ne: thanks for keeping up with the oslo changes. 16:38:36 thingee: Cool! Thank you to e0ne for is work on that! 16:38:40 jungleboyj: are we done with oslo_ changes after this? 16:38:55 e0ne: Thanks for keeping me honest wiht the changes. 16:39:08 jungleboy: :) 16:39:12 * thingee thinks he asks jungleboyj this once a week 16:39:30 We are mostly done with name space changes. I have the removal of config patch I am working on to get that out of incubator. 16:39:46 I have to get the scheduler workingt. 16:39:50 doh ok 16:39:52 jungleboyj: cool 16:40:14 so how much more do we want to get in then for K? 16:40:15 Once the logging change is in and the other two I will have a couple more clean-up patches. 16:40:19 those seem like big changes 16:40:21 late 16:40:40 thingee: As far as oslo_ goes. I think we are done. Looked a bit yesterday. 16:40:40 the oslo updates have bit us in the rear end a few times before late in the release cycles 16:40:54 so maybe early L 16:41:01 hemna: Right, that is why I want to get these wrapped up in the next week. 16:41:05 I should be able to do so. 16:41:05 thingee, +1 16:41:17 ok 16:41:24 hemna: The only big change is the scheduler part. 16:41:33 yah, that one, should wait IMHO 16:41:50 hemna might have a point... big changes at this point are high risk 16:41:52 hemna: I agree. I'm mostly not excited about oslo_log changes personally because I don't know what it's buying us to have it so late. 16:41:53 hemna: but if majority of the community is really wanting this and we can have it early after freeze, fine 16:41:53 hemna: re oslo_log 16:41:53 thingee: winston-d, what do you think? 16:42:11 logging is a small change, just hits an annoying number of files 16:42:12 thingee, +1 16:42:20 DuncanT: +1 16:42:23 DuncanT: +1 16:42:25 jungleboyj: again I'm not sure why we want oslo_log right away. So I'm unlikely to agree with anything else being updated at this point. 16:42:30 scheduler makes me more nervous 16:42:31 jungleboyj: the normalize support in scheduler filter? 16:42:36 yah the log patch is fairly limited. as long as oslo_log isn't broken itself, it should be ok. 16:42:40 winston-d: Yes. 16:42:44 8 days + whatever RC time for gate testing. I think we need to be careful at this point 16:42:53 thingee, +2!! 16:43:11 jungleboyj: don't we have a dependenecy in olso-incubator? 16:43:27 winston-d: There is a fix I would need to get in there. 16:43:31 jungleboyj: to add support for 'infinite' in normalize() 16:43:47 winston-d: Right. I am ok with waiting on that if that is ok with you. 16:43:58 does anyone disagree on waiting for things besides oslo_log? 16:44:03 in Kilo 16:44:19 thingee: Let me make the following proposal: 16:44:37 1) Get oslo_log change in for Kilo. 16:44:47 jungleboyj: please wait for a proper fix to accommadate 'infinite' 16:44:56 2) Hold off on scheduler change for Liberty. 16:45:12 just fyi, oslo feature freaze is on March, 12th 16:45:25 3) I move to the new config engine this week as it lets us get other cleanup in incubator done. 16:45:37 Remove files that we don't need to be carrying along. 16:45:58 Those will just be removing files we already have stopped using. 16:46:07 Then I will be done with Oslo. 16:46:15 1 and 2 seem fine. Need to see what 3 looks like 16:46:19 Config engine change makes me a little nervous, but gate should test it fairly well 16:46:41 DuncanT: +1 It shouldn't be a huge deal. 16:46:51 I will get it coded up and see what you guys think. 16:46:51 Need to follow up on any 3rd party failures though, since they are testing more of the config code than jenkins 16:46:52 jungleboyj: will it impact any package mainteiners? 16:47:04 e0ne: Not if I write it properly. :-) 16:47:20 jungleboyj: fair enough :) 16:47:25 jungleboyj: anything else? we have a couple more things on the agenda 16:47:43 e0ne: DuncanT thingee I will stop work on the scheduler and move to the config work so you have something to see yet this wek. 16:47:49 We can revisit next week. 16:47:56 jungleboyj: ok 16:47:59 ok 16:48:04 #topic Replication follow up from last meeting 16:48:05 thingee: Thanks for the discussion and input! 16:48:05 jungleboyj: hi 16:48:11 :) 16:48:13 thingee: He. 16:48:17 Hi. 16:48:20 :-) *sigh* 16:48:25 I can't type today. 16:48:30 jungleboyj: I'm going to do a hard cut on this in 6 mins. need to leave time for jbernard 16:48:31 Anyway, so I just wanted to follow up. 16:48:55 We took the discussion last week about GPFS and Storwize to heart. 16:49:02 So, this is a quick discusion. 16:49:26 We are not going to try to get any more replication changes in for this release and we are going to focus efforts on helping get V2 on track for Liberty. 16:49:41 jungleboyj: +1 16:49:44 You all are right that that makes more sense than trying to get more changes in now. 16:50:06 Hope that we can be of better benefit to everyone by focusing our efforts that way. 16:50:14 That was all I had on that. 16:50:17 :-) 16:50:43 I'll be providing more input on the replication stuff soon. Just in constant review mode :) 16:50:46 for kilo 16:50:58 thingee: Same here. 16:51:07 jungleboyj: thanks 16:51:10 We are continuing to look at the bugs out there. 16:51:26 9 minute warning 16:51:26 jungleboyj: I've been watching, thanks :) 16:51:30 #topic Generic volume migration consensus 16:51:33 jbernard: hi 16:51:35 hey guys 16:51:35 jgriffith: here? 16:51:42 thingee: Welcome. 16:51:53 my volume migraiton patch is availalbe for review https://review.openstack.org/#/c/135002/ 16:51:58 I've pinged jgriffith number of times to relook at this 16:52:03 winston has provided some great feedback 16:52:08 and we've iterated a few times 16:52:17 im left with two questions: 16:52:32 1. Do we agree that the patch in it's current form is not good enough to merge 16:52:49 2. If winston's suggested approach is implemented, is that what we want 16:53:09 and i guess also 16:53:19 3. we need CI for this, where does that fit in 16:54:01 the patch as it is, adds support for rbd migration wihtout introducing regressions for previous drivers 16:54:07 but the encapsulation is not awesome 16:54:50 and in general, im looking for some agreement so i don't go off in a wrong direction 16:54:51 * DuncanT would like to look at LVM + 3rd party driver for 3rd party CI in L. Probably not required initially, but it would allow at least some migration testing 16:55:11 also, we can revisit next week, i dont need an answer right now 16:55:18 it may take time to digest it all 16:55:23 but i wanted to bring it up 16:55:38 QED 16:55:56 * DuncanT didn't like the overloaded attach that doesn't attach, for the record 16:56:18 DuncanT: in the alternative? or as it is currently? 16:56:25 But I couldn't come up with anything better... am trying to make sense of winston's proposal now 16:56:29 ahh 16:56:45 me too, this morning has been busy 16:57:15 I hope avishay can sponse the review this generic vol migration patch. 16:57:31 ok, I don't have much input on this. Honestly haven't been following, so looking to others to chime in 16:57:33 he's the original author of migration. 16:57:45 3 mins 16:58:12 time for opens? 16:58:33 I have one 16:58:44 jbernard: thanks and sorry for the lack of discussion 16:58:49 #topic open discussion 16:58:53 Next week cinder meeting is Noon Eastern/9AM Pacific thanks to DST in the USA 16:58:58 no worries, if anyone does have time to look, i would be grateful 16:59:02 bswartz: +1 16:59:04 thanks for the reminder 16:59:18 Still the same time UTC, right? 16:59:23 make sure to send your summit session proposals here! https://etherpad.openstack.org/p/cinder-liberty-proposed-sessions 16:59:27 yeah UTC time never changes 16:59:28 bswartz: +1 Thank you! 16:59:55 1 minute 17:00:08 bug: https://bugs.launchpad.net/cinder/+bug/1406703 17:00:10 Launchpad bug 1406703 in Cinder "Deleting VM with an attached volume during copy-volume-to-image causes the volume remains in-use state" [Undecided,In progress] - Assigned to Abhijeet Malawade (abhijeet-malawade) 17:00:10 #endmeeting