14:00:12 #startmeeting glance 14:00:13 Meeting started Thu Jun 20 14:00:12 2019 UTC and is due to finish in 60 minutes. The chair is jokke_. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:14 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:16 The meeting name has been set to 'glance' 14:00:19 #topic roll-call 14:00:23 o/ 14:00:45 o/ 14:01:03 o/ 14:01:13 #link https://etherpad.openstack.org/p/glance-team-meeting-agenda 14:01:46 I think Brian is still off and Sean might be bit busy with his new job 14:02:25 yes, rosmaita will be on vacation till 24th 14:02:25 #topic release updates 14:02:52 Soo we got the glance_store sorted, thanks abhishekk for getting it out as well 14:03:08 \./ 14:03:24 so that release has fixed failing periodic jobs as well 14:03:31 now everything is green 14:04:44 yeap, so now we can move on, reintroduce that change and do all the dance around it 14:04:56 we are approx month away from Train milestone 2 release 14:04:59 and get the rest of the housekeeping done for 1.0.0 14:05:08 Yeah 14:05:18 jokke_, right 14:05:51 I will resubmit that patch tomorrow (backend >> store change) 14:06:46 sounds good. I went through pile or reviews yesterday ... small things that would be nice to tock out of the review queue 14:06:58 great 14:07:39 will go through another pile today 14:08:15 do we have anything else around release for now? 14:08:19 cool, I will followup with those tomorrow 14:08:26 that's it from me 14:08:50 #topic migration script for backend->store 14:09:23 yep, as we have discussed last week, I have added a script 14:10:02 kindly have a look, this is the best possible way to update the column value which has data type as PickleType 14:13:12 abhishekk: ok will have a look, obviously that is needed only after we do the key change ... thanks for having it ready to go 14:13:15 The query is compatible with sqlite and mysql as ell 14:13:22 nice 14:13:36 ack 14:13:48 s/ell/well 14:14:03 that's it from me 14:14:13 kk 14:14:19 #topic open discussion 14:14:57 I do not have anything at this time 14:15:08 I have one question, on cache enhancement we have decided to add wrapper around glanceclient 14:15:39 have we? 14:15:47 so is it target for this cycle or we can enhance cache-management now and add wrapper during next cycle? 14:15:48 * jokke_ is hitting totally blank on this 14:16:38 #link https://etherpad.openstack.org/p/Glance-Train-PTG-planning 14:16:42 line #57 14:17:05 oh yeah, abhishekk https://review.opendev.org/#/c/643336 could do with some love as well ;) 14:17:46 jokke_, will do shortly 14:19:35 ohh that one, correct me if I remember this totally wrong, but the idea was to wrap glanceclient once we get the caching stuff to API so that we can still have the glance-cache-manage for people's tools that exists atm. 14:21:14 jokke_, yes 14:21:37 so I have a PoC ready where I have moved caching stuff to API 14:21:50 with pre-caching of image 14:22:38 so at the moment I am wondering how can I keep all commands under same tool 14:23:44 if you send me patch file I'll have a look and bounce some ideas 14:24:03 or should I simply write new modules and keep old cache-manage tool as it is for the time being? 14:24:25 jokke_, I will send after the meeting 14:24:49 That's what I've been thinking but honestly I'm not 100% sure how much we need to blow up to migrate this stuff 14:25:13 that's why I thought I could have a look on your approach and bounce some ideas based on that 14:25:42 I would do new modules and make calls to them from the existing code base 14:25:56 keep it modular 14:26:45 davee_: so the problem is that the caching as it is now, is running inmemory sqlite db for the local cache management. Which the tooling and the middleware pokes 14:27:12 well it's inmemory that gets written eventually to disk for restarts etc. 14:27:28 and it's in each api-node 14:27:31 ahh, I need to get more familiar with ALL of the moving parts 14:28:36 yeah, the caching part has way more moving parts atm. than what would be great. Also most of the cache tooling is actually individual pieces of code that were designed to be ran independently 14:30:51 So one problem of single tool just interfacing the API of any given node is that it would need (admin) credentials to glance, which has quite a bit of security concerns vs. just having access to the local sqlite "directly" on the node 14:31:05 jokke_, just sent you the patch file (NOTE: as it is a POC I have removed some existing commands like queue, delete all etc) 14:31:53 abhishekk: sure ... I'll just have a look of your approach and see what we can get out of it ... I wasn't expecting it to be exactly production ready code even your pocs tend to be close normally :P 14:32:07 :D 14:32:19 salute 14:32:37 ./ 14:32:44 as I want to make sure we can reasonably integrate the Rabbit stuff to it as well 14:32:59 yeah 14:33:02 which is giving me headache ... surprisingly again due to the fecking onion 14:33:35 aha 14:34:38 you can imagine how much headache it is causing when I try to serialize our request to nice lil json blop so it can be sent to the message queue and deserialized on the receiving end for processing 14:34:58 and none of our damn objects are behaving as expected 14:34:59 Jokke: take 2 aspirins and call the team back in the morning, lol 14:35:29 Its really a nightmare 14:36:32 abhishekk: I might need to be doing it literally at the top API layer just below mapping to keep it out of the onion 14:37:12 jokke_, yes, that would be better 14:37:21 I would like to have a look at the design if possible 14:37:51 like catching the response we'd be sending, if it's something else than 2XX then sending out the MQ fanout and seeing if we get 2XX from there 14:38:07 and reconstructing the response 14:39:14 just lots of pokeypokey 14:39:45 and obviously blocking calls keeping the client waiting 14:39:55 right 14:41:02 but I keep poking it and hopefully your approach on the caching gets me some inspiration how we can streamline at least some parts of it 14:41:36 I hope so that my approach will make sense :P 14:41:58 btw, I'll be working from Finland for the beginning of July so earlier TZ 14:42:27 cool, thank you for the update 14:42:55 I will might take a leave or two next wee due to festival 14:44:55 ah, something going on locally again? 14:45:18 that's not the big festives is it? 14:45:44 nope, we will have big one in september and october end 14:46:19 ah it was the last fall denver ptg we were talking about that not the one at the spring :) 14:46:46 ohh you remember :D 14:47:08 ofc 14:47:31 well I think that's all from me, reviews would be welcome, lets try to clean out our Q a bit 14:47:39 ack 14:47:50 nothing from me as well 14:47:54 I'll hopefully have better update for the Rabbit stuff next week 14:48:18 great, all the best with onion :D 14:48:19 I'll see if I can catch up with you abhishekk tomorrow morning about the cache stuff 14:48:26 I'll have a look at it tonight 14:48:31 jokke_, that would be nice 14:48:37 thank you 14:48:55 ok, thanks all! getting your 10min this afternoon back :) 14:48:57 thank you all, have a good day ahead 14:49:03 #endmeeting