15:01:46 <bswartz> #startmeeting manila 15:01:47 <openstack> Meeting started Thu Dec 22 15:01:46 2016 UTC and is due to finish in 60 minutes. The chair is bswartz. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:48 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:49 <bswartz> hello all 15:01:51 <tommylikehu_> Merry xmas 15:01:51 <openstack> The meeting name has been set to 'manila' 15:01:51 <cknight> Hi 15:01:56 <vponomaryov> hello 15:01:58 <tommylikehu_> hi 15:01:59 <ganso> hi 15:02:02 <tbarron> hi 15:02:04 <ravichandran> hi 15:02:09 <markstur> hi 15:02:16 <xyang1> hi 15:02:32 <bswartz> #topic announcements 15:02:46 <bswartz> We passed the driver proposal freeze this monday 15:02:58 <gouthamr> hello o/ 15:02:59 <bswartz> it looks like we have 3 new drivers to consider this cycle 15:03:52 <bswartz> also due to the holidays I think we should cancel the meeting next week 15:04:05 <bswartz> I know at least 5 cores will be on vacation 15:04:15 <bswartz> and I'm guessing the rest will probably be too 15:04:19 <xyang1> I am on vacation too 15:04:31 <tommylikehu_> cool 15:04:31 <bswartz> okay so it's an easy decision 15:04:41 <bswartz> #info Dec 29 meeting cancelled 15:05:11 <bswartz> Next meeting should be Jan 5 then 15:06:01 <bswartz> I know some people will still be on vacation that week but with FPF coming up Jan 9 we need to meet that week 15:06:25 <bswartz> #agenda https://wiki.openstack.org/wiki/Manila/Meetings 15:06:31 <bswartz> #topic specs implementaion status 15:06:35 <ganso> bswartz: isn't FPF Jan 12th? 15:06:38 <vkmc> o/ 15:06:42 <vkmc> hi hi 15:06:51 <bswartz> ganso "week of" Jan 9th 15:06:55 <ganso> bswartz: oh ok 15:07:31 <bswartz> personally I'd prefer we do freezes on mondays instead of thursdays but I didn't announce an exact date this cycle 15:07:51 <bswartz> so it can be thursday 15:08:11 <bswartz> okay let's talk about specs 15:08:24 <bswartz> thankfully we got our specs straightened out last week due to a last minute push 15:08:41 <bswartz> the bad news is that we have almost no time to review the implementations 15:08:55 <bswartz> and implementation of some specs is blocked by dependencies 15:09:27 <bswartz> the goal of the specs process was to limit what we focus on to conserve review time, but we're in danger of still having too much code to review 15:09:56 <bswartz> please prioritize review of high priority spec implementations over low priority ones 15:10:50 <bswartz> #link https://github.com/openstack/manila-specs/blob/master/priorities/ocata-priorities.rst 15:10:55 <bswartz> THOSE ARE HIGH PRIORITY SPECS 15:11:00 <bswartz> HOLD ON CAPS LOCK IS STUCK 15:11:12 <ganso> lol 15:11:14 <bswartz> that's better 15:11:21 <markstur> :) 15:11:26 <vponomaryov> =) 15:12:11 <bswartz> vponomaryov: do you have any changes related to the scenario tests specs that are waiting? 15:12:19 <vponomaryov> bswartz; no 15:12:20 <xyang1> bswartz: have you submitted any patches for eliminate race conditions? 15:12:31 <bswartz> xyang1: that one is next 15:12:35 <vponomaryov> bswartz: both changes I had has been merged 15:12:37 <xyang1> ok 15:12:51 <vponomaryov> s/has/have/ 15:12:54 <bswartz> vponomaryov: so everything for ocata is done? 15:13:10 <vponomaryov> bswartz: according to time ranges, Iguess yes 15:13:17 <bswartz> okay 15:13:30 <bswartz> regarding race conditions, this is the next patch that needs to land: https://review.openstack.org/#/c/318336/ 15:13:42 <bswartz> I'm working on that review 15:14:14 <bswartz> we need some others with tooz experience (other than tbarron and gouthamr, who are authors on that change) 15:14:32 <xyang1> do we have an etherpad that lists all the code patches associated with high priority specs? 15:14:46 <bswartz> xyang1: no, would that be helpful? 15:14:51 <xyang1> yes 15:15:07 <bswartz> do we already have an etherpad we can reuse? 15:15:11 * bswartz goes looking 15:15:30 <markstur> https://etherpad.openstack.org/p/manila-ocata-code-review-focus 15:15:58 <xyang1> we could add code patches to this spec 15:16:07 <tommylikehu_> ok 15:16:16 <bswartz> awesome 15:16:33 <tommylikehu_> it's not the high priority ones 15:17:07 <bswartz> yeah I need to add the specst hat merged last week at the top of this list 15:17:15 <bswartz> I won't take up our meeting time doing that though 15:17:57 <bswartz> okay access rules 15:18:03 <bswartz> gouthamr: any patches waiting for review? 15:18:27 <gouthamr> bswartz: nope.. i haven't pushed one of the patches and the parent patch needs an update too 15:18:44 <gouthamr> bswartz: so, not ready yet. hoping to get it in soon 15:18:48 <bswartz> okay and ipv6 is on my list for review 15:18:51 * bswartz finds link 15:19:26 <bswartz> #link https://review.openstack.org/#/c/312321/ 15:19:44 <ganso> bswartz: are we going to discuss spec details in this meeting? 15:20:02 <bswartz> ganso: there shouldn't be any details to discuss... 15:20:12 <bswartz> ganso: oh you're referring to the proposed changes to mountable snapshots 15:21:01 <ganso> bswartz: not only that, there is one detail I don't fully agree in the access rules spec, we could discuss the possibility of a change 15:21:13 <ganso> bswartz: there was some discussion in the channel yesterday 15:22:02 <ganso> bswartz: we could come back to it after going through all the specs 15:22:04 <bswartz> okay well one of the goals of the specs process was to get all the design decisions out of the way before the spec merged -- so in theory we should leave the spec the way it is 15:22:14 <bswartz> however we do make mistakes 15:23:30 <bswartz> ganso: okay we can come back to that topic 15:23:40 <bswartz> regarding ipv6, any other patches looking for review? 15:24:16 <tommylikehu_> sure 15:24:32 <tommylikehu_> I will add these in the etherpad 15:25:38 <bswartz> okay 15:25:42 <bswartz> we can move on then 15:26:06 <bswartz> #topic 24 hour rule 15:26:11 <bswartz> tbarron: you added this? 15:26:22 <tbarron> Recently I was reminded of the manila "24 hour agreement" w.r.t. +A workflow. I either hadn't been told it or forgot it, and it seems other cores were in my same boat, so it seemed we should briefly review it here. 15:26:37 <tbarron> just a quick statement for common understanding 15:26:52 <tbarron> vponomaryov: ^^ 15:27:15 <gouthamr> 24 hour agreement? 15:27:22 <bswartz> this agreement isn't just for manila it's across the whole community right? 15:27:27 <xyang1> I have not heard about its before 15:27:29 <tommylikehu_> what's that 15:27:33 <ravichandran> whats that 15:27:53 <bswartz> maybe the 24-hour rule only exists in a few people's heads 15:27:55 <markstur> it's from vponomaryov's secret club rules 15:27:56 <vponomaryov> long time ago we agreed to not merge changes before 24 hours pass after its uploading 15:27:56 <tbarron> bswartz: seems to be something old-timers know and new-timers don't 15:28:24 <vponomaryov> with exception of merging blocker fixes 15:28:34 <cknight> tbarron: so you're saying you're a new-timer? 15:28:35 <xyang1> I don't think I am that new:) 15:28:41 <vponomaryov> the goal - allow people from various time zones to react 15:28:48 <tbarron> cknight: and old-old-timers forget 15:28:54 <cknight> xyang1: You're newer than tbarron :-) 15:29:04 <xyang1> am I?:) 15:29:21 <bswartz> yeah this is one of those "common courtesy" rules which is that since many of us are on strange time zones, in order to give everyone a chance to at least look at new patches we don't merge new things in less than 24 hours 15:29:29 <tbarron> xyang1: in OpenStack years you are my elder :D 15:29:35 <vkmc> is there documentation wrt reviewing in Manila (including this rules)? 15:29:40 <vkmc> maybe in some wiki 15:29:42 <tommylikehu_> 24-hours delay? 15:29:48 <xyang1> tbarron :) 15:29:56 <bswartz> it's mainly to avoid stuff slipping by, except for gate fixes 15:30:15 <ganso> is it for changes or patchsets? 15:30:27 <xyang1> I remember long time ago eharney had said something like this in Cinder 15:30:28 <bswartz> it's mostly about making sure everyone has a chance to look at changes if they want to 15:30:31 <vponomaryov> ganso: it is for "+A" button 15:30:49 <ganso> if it is for changes, a new patchset that changes drastically the code could get merged 15:30:51 <vponomaryov> ganso: and mostly was about commit 15:30:54 <gouthamr> only seen trivial fixes being merged within a short time.. is this a new concern here? 15:30:57 <bswartz> vponomaryov: I think the question is what about a change that's been up for a while, and then a new patchset comes in 15:31:20 <bswartz> are you expecting us to wait 24 hours after the last patchset? or just 24 hours since the change was created? 15:31:34 <gouthamr> i mean TrivialFixes usually merge within 1 or two patchsets, and without having to pass all third party CIs 15:31:43 <vponomaryov> bswartz: personally, I consider diff with previous PSs 15:31:54 <vponomaryov> bswartz: for other reviewers and my own 15:32:11 <bswartz> because lots of times a change has been up for a while, then someone reviews it and -1 and the author addresses the comments right away 15:32:30 <bswartz> then asks for +2 based on changes 15:32:31 <xyang1> vponomaryov: do you have a specific example that the patch was merged too quickly? I think this could depends on what the patch is about 15:33:04 <vponomaryov> xyang1: recently - one of commits merged by tbarron 15:33:12 <vponomaryov> xyang1: only by him 15:33:24 <bswartz> ninja'd? 15:33:27 <vponomaryov> tbarron: no blame, just fact ) 15:33:37 <ganso> vponomaryov: but that is breaking another rule mostly 15:33:39 <markstur> he named names 15:34:04 <gouthamr> okay tbarron, no gifts for you this year 15:34:05 <markstur> but this is how we learned about the mythical 24h rule 15:34:27 <vponomaryov> gouthamr: maybe he is the one who presents gifts? )) 15:34:28 <ganso> we may need to call the OpenStack Police Department 15:34:37 <tbarron> I don't remember the change, it didn't seem a big deal at the time :D 15:34:38 <markstur> vponomaryov: LOL 15:34:53 * gouthamr tbarron is the one who knocks 15:34:55 <tbarron> but I have not problem following the rule if we all know what it is 15:35:04 <vponomaryov> tbarron: it was the hand of destiny, now everyone know the RULE 15:35:11 <bswartz> tbarron: personally I think some judgement is needed 15:35:28 <bswartz> there are probably changes that this rule doesn't need to apply do, and some where it's VERY IMPORTANT 15:35:41 <xyang1> bswartz: +1 15:35:44 <vkmc> ganso, lol 15:35:48 <markstur> vponomaryov: can you re-arrange the letters in tbarron to spell santa? No that'd be too obvious. 15:35:55 <xyang1> bswartz: I think it really depends on what change it is 15:36:08 <tbarron> as a matter of common sense I don't do +2+A on stuff that I think others should examine. 15:36:30 <vponomaryov> ok, everyone disagree with t he rule? 15:36:39 <vponomaryov> /disagree/agree/ =) 15:36:43 <markstur> I think it is good to talk about 15:36:43 <bswartz> the key thing to remember is that some people wake up every day and look at all the new patches, and if you merge something in less than 24 hours you might be denying those people a chance to register their opinion 15:37:07 <xyang1> vponomaryov: I think we have to see what patch it is 15:37:10 <bswartz> so our of courtesy to your fellow cores, don't be in a hurry to workflow new things 15:37:17 <gouthamr> +1 -> applies to TrivialFix as well? 15:37:27 <xyang1> vponomaryov: whether it is controversial or a simple change 15:37:51 <vponomaryov> gouthamr: it happens that commit called trivial is not indeed 15:37:56 <bswartz> this isn't a written-down rules it's more about courtesy 15:38:07 <bswartz> and I agree that "trivial" is a judgement call we might not agree on 15:38:30 <ganso> vponomaryov: if it is not Trivial it shouldn't have been merged as Trivial, the 24h rule is not the problem here 15:38:57 <vponomaryov> ganso: if we have 2 +2 made by mistake in 1 hour? 15:39:07 <vponomaryov> ganso: then it is 15:39:16 <ganso> vponomaryov: which patch is it 15:39:17 <vkmc> IMHO there is place for a lot of subjectivity to play in with this... what is considered trivial here? 15:39:36 <vkmc> why you directly don't impose something like... 2 +2 are needed and then the WF+1 should be done by a different core? 15:39:47 <bswartz> vkmc: the main criteria for a trivial fix is that everyone agrees that it's trivial 15:40:03 <markstur> or just tbarron 15:40:09 <bswartz> vkmc: I don't want to make our rules more complicated 15:40:12 <vponomaryov> vkmc: and to be able to agree or disagree I should catch this train first 15:40:14 <vkmc> bswartz, +1 15:40:26 <ganso> vkmc: that's already imposed 15:40:48 <ganso> vkmc: oh nevermind the different WF +1 is not imposed 15:41:02 <vkmc> ganso, np 15:41:08 <bswartz> okay so have we covered this topic enough? 15:41:10 <vponomaryov> and little clarification 15:41:31 <vponomaryov> not 2 +2, but 2 +2 from 2 different companies 15:41:37 <vponomaryov> at least 15:41:47 <tbarron> yeah, that one I know :D 15:41:49 <vkmc> yup, I knew about that rule :) 15:41:52 <bswartz> yeah that restriction only applies to me, cknight, and gouthamr currently 15:42:28 <bswartz> to prevent netapp from weilding unchecked power 15:42:59 <bswartz> okay I see one other topic was added to the agenda 15:43:06 <bswartz> #topic Manila CI Tempest Test Cases are Failing 15:43:14 <bswartz> ravichandran: what did you want to say about this? 15:43:15 <ravichandran> I am observing few CI tempest test cases are failing randomly . Just want to know whether anything changed in last 2 to 3 weeks related CI which requires individual driver changes. 15:44:26 <vponomaryov> ravichandran: which CI jobs are you talking about? 15:44:26 <bswartz> ravichandran: can you tell us which tests and/or link to some jobs that illustrate this? 15:44:40 <bswartz> stuff has definitely changed, but mostly fixes to make things better 15:44:48 <vponomaryov> ravichandran: we have lots of third-party CIs that fail 15:44:55 <ravichandran> HPE 15:45:31 <ravichandran> oh ok I will email the details 15:45:51 <vponomaryov> email? whom? 15:46:12 <vponomaryov> for such cases, just ask for help in manila chat 15:46:51 <xyang1> email to markstur:) 15:47:01 <ravichandran> :) 15:47:10 <markstur> xyang1: hey! 15:47:12 <vponomaryov> xyang1: he has IBM badge now )) 15:47:12 <bswartz> lol 15:47:18 <xyang1> :) 15:47:35 <bswartz> you can quit HP but you can never leave 15:47:51 <markstur> There was a devstack CI breakage that folks were fixing. I have a note that says see IRC logs Dec 6. Checking. 15:47:57 <vponomaryov> HPkudza? ) 15:48:18 * markstur is wearing an HP hoodie and vest right now 15:48:52 <bswartz> http://i.kinja-img.com/gawker-media/image/upload/t3evzz4gbm7gql9u3x2t.gif 15:49:04 <vkmc> aahaha 15:49:22 <ravichandran> :) 15:49:30 <bswartz> okay hopefully we can get ravichandran's CI sorted out offline 15:49:33 <bswartz> #topic open discussion 15:49:47 <bswartz> anything else then for today? 15:50:01 <ganso> bswartz: are we going to discuss the access rules spec detail in the channel? 15:50:08 <bswartz> oh crap 15:50:12 <bswartz> yes I said we'd come back to that 15:50:19 <bswartz> let's cover it now 15:50:19 <markstur> ravichandran: http://eavesdrop.openstack.org/irclogs/%23openstack-manila/%23openstack-manila.2016-12-06.log.html 15:50:21 <bswartz> we have 10 minutes 15:50:40 <ganso> ok so, as some of you know, the mountable snapshot spec has a problem regarding access rules 15:50:53 <ganso> it was proposed with the old approach allow/deny driver_interfaces 15:50:56 <ganso> so I am reworking that 15:50:59 <markstur> ravichandran: If it broke back then. Some CIs had to move from "post-extra" to "test-config" 15:51:24 <ganso> while I was doing that, I noticed that the new spec proposed for access rules does not handle recovery mode anymore in scenarios other than driver restarts 15:52:00 <ganso> At Tokyo, one of our main motivations (and one of the main motivations for me to rewrite the old approach to the new update_access approach) is that we needed the recovery mode to fix invalid access rules in the backend 15:52:13 <gouthamr> markstur ravichandran: careful with that change though, you only need to do that on "master" and not stable branches 15:52:21 <bswartz> ganso: we have a long-pending topic about ensure_share and general cleanup actions 15:52:27 <ganso> in my opinion, restarting a service to fix access rules in a backend is too drastic 15:52:48 <ganso> bswartz: I think this is unrelated to ensure_share, at least considering the current scope of that 15:53:05 <bswartz> ganso: IMO it's important that driverse have the capability to correct access errors whether we use it or not 15:53:15 <tbarron> ganso: can you remind us why some backends need recovery mode? 15:53:23 <tbarron> ganso: and how it works? 15:53:31 <bswartz> ganso: mostly I'm thinking about when a manager service is killed while modifying an access rule 15:53:56 <bswartz> ganso: upon restarting, we should easily be able to detect that that happend 15:54:18 <bswartz> ganso: and it's important that we can roll-forward the pending access change 15:54:25 <ganso> tbarron: if a backend cannot add/delete an error-ed rule for some reason 15:55:19 <tbarron> ganso: so it flushes all rules and rebuilds from the DB? 15:55:22 <bswartz> ganso: or at the very least, when the *next* access rule change comes, it should cause the access rules for that share to be rationalized 15:55:36 <markstur> If I try to create an access rule and it fails. Then I have an error status. Then an attempt to delete it, does "recovery"/"resync". 15:55:42 <ganso> tbarron: yes 15:55:59 <tbarron> ganso: so we have two use cases then, yours and bswartz's 15:56:03 <bswartz> guys I'm not talking about cases where bugs cause the rule modifaction to fail outright 15:56:17 * gouthamr is in a standup meeting.. will react to scrollback in a bit 15:56:27 <bswartz> I'm talking about the case where we don't know if the operation completed or not because the service simply died while the thread was still down in the driver 15:56:56 <tbarron> bswartz: yeah, I agree that's a real use case. Also a driver might for its own reasons want a full refresh. 15:57:46 <ganso> tbarron: we add and remove rules while we don't exactly know what is in the backend, we believe it is always in sync with the manila DB, but it may run out of sync 15:57:47 <tbarron> with the new spec, the DB state is fully set in the API service. So full state (what is there, plus any pending operations) could be pushed to the driver. 15:58:09 <tbarron> ^^ when there is a need 15:58:14 <bswartz> yes that's the main point of the driver interface change 15:58:18 <tbarron> not for routine updates 15:58:27 <ganso> tbarron: yes, we need to define triggers for that. According to the spec, the only trigger is a driver restart 15:58:36 <ganso> tbarron: s/driver/service 15:58:40 <tbarron> driver could return a code to manager asking for a refresh. 15:58:44 <bswartz> if we always tell the driver the complete set of rules, and the driver is able to push those to the backend, it becomes impossible to get out of sync 15:59:17 <bswartz> the first if is easy -- we've done it 15:59:26 <bswartz> the second if depends on the backend APIs and the driver code 15:59:30 <ganso> bswartz: but the implementations follow one of two paths: they don't look at the full list if they receive a rule to add or delete 15:59:46 * vponomaryov thinks why we still don't have API to force access rule recheck... 15:59:50 <ganso> bswartz: so they don't know if they have run out of sync 16:00:28 <bswartz> ganso: we added an optimization path for drivers where it's extremely slow or expensive to get the current access rule list *cough* ceph *cough* 16:00:50 <gouthamr> time check 16:00:52 <bswartz> however all drivers are required to implement correct behavior 16:01:00 <bswartz> yeah we're past our time 16:01:07 <bswartz> ganso: did we answer your question? 16:01:12 <ganso> bswartz: not really 16:01:14 <gouthamr> is ready to answer ganso on #openstack-manila 16:01:24 <bswartz> okay let's head back to our channel and sort it out there 16:01:35 <bswartz> thanks all and have safe a happy holidays! 16:01:47 <bswartz> #endmeeting