14:01:38 #startmeeting glance 14:01:42 Meeting started Thu Feb 28 14:01:38 2019 UTC and is due to finish in 60 minutes. The chair is jokke_. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:01:43 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:01:46 The meeting name has been set to 'glance' 14:01:47 #topic roll-call 14:01:49 o/ 14:02:36 o/ 14:03:49 #link https://etherpad.openstack.org/p/glance-team-meeting-agenda 14:03:51 o/ 14:03:58 oh there's Brian 14:04:27 got here earlier than i expected, no traffic! 14:04:35 great 14:04:45 I'm dualboxing this meeting slot so sorry if I'm bit distracted 14:04:50 o/ 14:05:21 hi smcginnis 14:05:25 so lets kick off 14:05:45 I have no big announcements here so lets go with abhishekk 14:05:50 #topic release updates 14:05:57 ack 14:06:20 glance_store release tomorrow 14:06:25 Just want to make quick note vs. the agenda. We do not release Fridays so the store release day is today 14:06:27 all summary is here 14:06:43 #link https://etherpad.openstack.org/p/glance-stein-milestone-3 14:07:09 jokke_, ack 14:07:13 Thanks jokke_. Release milestone deadlines are always Thursdays. 14:07:35 abhishekk: that's purely to make sure if we break the world we don't need to spend our weekend fixing it 14:07:37 I have a release patch ready, just need to add hash there 14:08:00 jokke_, :D, got it 14:08:37 what's the status of the stuff on the etherpad? 14:08:44 so that comes down to question, are we allowing Rethinking File System to get in? 14:09:27 rosmaita, https://review.openstack.org/#/c/533413/, need to take decision on this 14:09:38 and this, https://review.openstack.org/639765 14:09:53 last link is about rethinking filesystem access patch 14:10:37 so it's time to decide whether we are willing to get this change in last minute, even though I am sure it will not cause any regression at the moment 14:10:55 So my opinion like said last week still stays, as we're not moving multiple backends from EXPERIMENTAL this cycle we should not utilize it in those places either as that would need the backends to be configured via the multiple backends interface 14:12:01 ok, that was my question, what we were doing about experimental backends 14:12:31 i mean, the multiple backends feature being experimental 14:13:12 Yes, I thought that was agreed already month ago or so 14:13:15 apart from these we need release-notes patch for glance-store as well 14:13:28 abhishekk: I could help with that 14:13:35 jokke_, great 14:14:25 So my main concern with the multiple backends is how we deal with the fact that consumers, mainly nova and cinder are not multiple backend aware and we need to come up with proper solution about that 14:14:45 good forum topic! 14:14:51 jokke_, I am planning to submit a spec in nova and cinder for the same 14:15:08 I will do the same by next week (mostly) 14:15:10 for stuff like the locations (and thus Nova ceph snapshots) are not targeted 14:15:39 rosmaita, ++ 14:15:45 abhishekk: great ... and we need to try to fit that discussion to PTG as common topic between us and Nova&Cinder 14:16:04 jokke_, ++ 14:16:15 yeah, i think what we need is a clear, well-publicized default strategy 14:16:15 And like said, I'm booking half day of PTG purely for sorting this out 14:16:29 Good plan. 14:16:32 cool 14:16:54 This is important and big change and we need to get it right, preferably first time :D 14:17:31 and it will be :D 14:18:08 Ok so glance_store only thing we need for the release is the renos then? 14:18:36 so again are we planning to get rethinking file store patch in, or move it to Train? 14:18:57 15:04 < jokke_> So my opinion like said last week still stays, as we're not moving multiple backends from EXPERIMENTAL this cycle we should not utilize it in those places either as that would need the backends to be configured via the multiple backends interface 14:19:11 Any other opinions ^^ 14:19:16 Ok, so basically hold off on rethinking filesystem for Train 14:19:36 rosmaita, thank you :D 14:19:51 I was lost in reading this entire big statement 14:20:04 So if I have understood our backend code correctly, doing that now would force the users to move to the experimental multiple backends 14:20:10 in their whole deployment 14:20:16 we can't use it only for those 14:20:41 I will add workflow -1 to both the patches then 14:20:47 with message 14:21:22 abhishekk: cool, I've been keeping eye on those patches and IMO that will be very easy merge as soon as we open Train development 14:21:34 the work you've been doing looks great 14:21:42 i think we are being super-conservative here 14:21:43 jokke_, yes 14:21:57 the not making people use multiple backends was for the interface config 14:22:00 Will probably be good to have the full cycle to shake out any unintended side effects or things we didn't think of. 14:22:08 i think we could use them "internally" just fine 14:22:20 but ... it's probably too late to introduce that change 14:22:28 rosmaita: If I understand correctly, the store code works in a way that you can't use both simultanously 14:22:46 it will barf if you have old and new way configured together 14:23:10 gotcha 14:23:14 So if we use them internally no-one can configure the stores the old way anymore 14:23:18 jokke_, the recent change I have made will work only if multiple backend is enabled 14:23:25 ^^ 14:23:41 i guess the safest thing is to just hold off for Train 14:23:44 jokke_, it will not have any side-effect if stores are configured using old way 14:23:58 we can call it the A-Train release (A for Abhishek!) 14:24:10 :) 14:24:10 rosmaita, agree, it's very late to add this now 14:24:12 :D 14:24:15 abhishekk: are you saying that we could use them both parallel? 14:24:23 jokke_, yes 14:24:36 Feck, sorry. I have totally missed that 14:24:53 that has been literally my only concern not pushing those for Stein 14:25:16 as the whole design of multiple backends was that if you use them, you must use them 14:25:40 I am saying if enabled_backend is defined then reserved stores will be injected while service starts and if enabled_backend is not defined then it will use node_staging_dir and work_dir will be used as it is 14:25:43 And that's what I said last week as well :( 14:25:51 well, since glance_store must go out today, i think if there's any question, we need to not release 14:26:02 i mean , not include the rethinking filesystem 14:26:17 rosmaita, I guess we should move it to Train 14:27:08 The only patch in question is now, https://review.openstack.org/#/c/533413/ 14:27:08 the store change is 10 lines 14:27:25 can we review it literally now and leave the door open for that? 14:28:21 #topic rethinking filesystem store usage 14:28:26 jokke_, I am ok to move it to next cycle 14:28:50 #link https://review.openstack.org/#/c/639765 14:29:05 (I have already added workflow -1 to it :D) 14:30:00 I'm ok either way, just sad that it was blocked due to misunderstanding 14:30:14 you know, there's no reason why we can't cut RC-1 early to open up train developmnet 14:30:28 assuming we've got most of the stuff we want in there 14:31:04 rosmaita: yeah, well we have time on glance side, it's the 10 line store change that would have needed to be in the store release 14:31:52 I guess lets move ahead, and In T we will able to expand this to use other stores as well? 14:32:42 The benefit of that is that we will have stable stamped feature set to use in that 14:33:00 ok 14:33:58 So only release-note and https://review.openstack.org/#/c/639765 decision on this for store release 14:34:22 I am ok to get this patch in, jokke_ has -1 on it 14:35:01 639765 I don't have -1 it's that your patch 14:35:27 sorry, https://review.openstack.org/#/c/639765 14:35:35 same link ;) 14:35:44 https://review.openstack.org/#/c/533413/ 14:35:50 this time it is correct :D 14:36:05 yes 14:36:12 I am ok to get this in 14:36:29 jokke_ had concerns about this one 14:36:44 yes 14:36:55 So I hate that change in principle level. If we really want to do that, technically the patch looks good 14:37:36 it's really simple change that will just be counter productive for lots of the work we have been doing in sense of streamlining and making reasonable defaults 14:38:36 I am ok to have this in, but looking at gate how much time it will take to get this in? 14:38:53 i don't know enough about the ceph backend tbh to know whether this is good or bad 14:39:14 I have release patch ready and will be around for next 4-5 hours 14:39:16 maybe let's put this on as a forum topic and see what operators want 14:39:26 rosmaita, +1 14:39:26 and not include it now 14:39:48 good, then we require release-notes patch only now 14:40:03 Ok, so release notes and release are the only things we need to take care of today then 14:40:13 \o/ 14:40:50 I will format the etherpad tomorrow, so that it will give us clear idea about glance patches with it's current status 14:41:01 #topic when to create the folders for Train Specs 14:41:13 what is the ideal timeline for this? 14:41:16 I'd say we can do this next Fri 14:41:21 probably when S feature freeze happens 14:41:32 as in once the milestone 3 is past 14:41:39 or FF either is fine by me 14:41:56 rosmaita, jokke_ which means we can do it tomorrow or next friday 14:42:03 I wouldn't mind to leave nice gap before summit/ptg for people to sumbit specs in 14:42:23 abhishekk: milestone 3 is week from now, not today 14:42:24 I will put a patch tomorrow 14:42:40 FF is I think 2 weeks after that 14:42:52 jokke_, ohhh 14:43:21 it's just the non-client lib deadline that is today 14:43:24 FF is one week after that 14:43:49 yeah, but since we already had spec freeze, i think we can open the specs repo 14:44:05 ok 14:44:16 that's it from me 14:44:37 ok, lets get it done next week with M-3 14:44:50 and we're somewhat consistent with previous cycles 14:45:22 ok 14:45:25 #topic Open Discussion 14:45:29 Anything else? 14:45:38 nope, 14:46:22 rosmaita: smcginnis: ? 14:46:29 Nothing from me. 14:46:31 ohh regarding periodic jobs, subunit issue is still there but it's less frequent now only 5 failures in last week, 1 in last two days 14:46:54 Oh great. 14:47:20 abhishekk: that is good news indeed 14:47:33 (not that it's there but the frequency is reduced 14:47:35 sorry, was distracted 14:47:54 (I will be back around 2-3 hours and will put a release patch up after that) 14:48:07 rosmaita: np, just making sure we're not missing anything from you before I close 14:48:19 abhishekk: do you want me to do the release so you don't have to come back? 14:48:59 rosmaita, I need to travel back so once I reach home and release-notes patch is not merged by then I will ping you 14:49:04 rosmaita, thank you :D 14:49:16 ok, i will be afk for an hour, but online for hte rest of my day after that 14:49:26 cool 14:49:28 no more topics from me 14:49:32 cool 14:49:44 thank you all 14:49:47 thanks! 14:49:54 thanks eveyone! 14:49:57 #endmeeting