14:00:42 <jokke_> #startmeeting glance
14:00:44 <openstack> Meeting started Thu Aug  8 14:00:42 2019 UTC and is due to finish in 60 minutes.  The chair is jokke_. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:45 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:47 <openstack> The meeting name has been set to 'glance'
14:01:21 <jokke_> #topic roll-call
14:01:23 <jokke_> #link https://etherpad.openstack.org/p/glance-team-meeting-agenda
14:01:27 <jokke_> o/
14:01:48 <zhengMa> Hi
14:02:11 <rosmaita> o/
14:03:03 <jokke_> abhishekk said he might be bit late, we have very thin agenda for today so lets give couple of min for him and anyone else who wants to join to catch up with us
14:03:15 <rosmaita> sounds good
14:04:14 <rosmaita> jokke_: while you are waiting, take a look at this: https://review.opendev.org/#/c/674697/
14:05:11 <rosmaita> i did that instead of writing release notes
14:06:21 <jokke_> kk, have it open
14:06:26 <jokke_> lets get started
14:06:34 <jokke_> #topic updates
14:07:25 <jokke_> so first of all, I'm back and try to go through tons of reviews over next few days. We have lots of important stuff in the queue
14:08:02 <jokke_> If there is something that seems like I have missed by mid next week, feel free to ping me for attention
14:08:57 <jokke_> Obviously the cycle is running out quickly as always, but I feel like we have got lots done (namely the g_s 1.0.0 and multistore) and we have loads in flight
14:08:59 <rosmaita> ok, i take it you are working through abhishek's m-2 etherpad?
14:09:32 <jokke_> I will look that as well, but I'm literally going to work through couple of pages of our review queue
14:10:07 <abhishekk> o/
14:10:16 <jokke_> just to make sure we catch some of the housekeeping stuff as well what people have submitted
14:10:19 <jokke_> hey abhishekk
14:10:34 <abhishekk> hi, sorry for late :D
14:10:38 <jokke_> np
14:11:49 <jokke_> I will be in the Shanghai summit (assuming that I will get visa, should know in couple of weeks) so Glance PTG is go and will see if we can get some forum stuff in as well
14:12:17 <abhishekk> I will be there too, assuming visa is granted
14:12:52 <jokke_> I think rosmaita and abhishekk are coming too, right?
14:13:02 <rosmaita> yes
14:13:10 <abhishekk> great
14:14:19 <jokke_> great, so we will have good representation, lets make sure we get best out of that time. Also for any of our community members ini APAC who might not have been able to join to the summits/PTGs before and will/can be in Shanghai, I'm looking to meet ye all!
14:15:03 <jokke_> Remember that the early bird ticketing is ending at 14th of Aug
14:15:30 <jokke_> so next week, and the ATC code that should have dropped into your e-mail gives 30% off from the price
14:16:11 <jokke_> if anyone has problems, like not received ATC code, I'm more than happy to try to help sorting those out with the foundation
14:16:12 <abhishekk> I have already purchased :D
14:16:44 <jokke_> ok, that's all from me, giving the mic to abhishekk
14:16:50 <jokke_> #topic release updates
14:17:21 <abhishekk> Regarding periodic jobs everything is green at the moment
14:17:47 <abhishekk> Just for update we had store 1.0 out two weeks before
14:17:59 <davee_> :)
14:17:59 <abhishekk> #link https://etherpad.openstack.org/p/Glance-Train-MileStone-2-Release-Plan
14:18:20 <abhishekk> Above etherpad contains all the important patches for glance
14:18:28 <abhishekk> Kindly have a look and review
14:18:58 <abhishekk> Most of the patches are reviewed by rosmaita and sean
14:19:09 <abhishekk> Thank you rosmaita and sean for your help
14:19:14 <rosmaita> np
14:19:18 <abhishekk> that's it from me
14:19:40 <jokke_> ok
14:19:47 <jokke_> #topic Open Discussion
14:20:32 <jokke_> First of all, big thanks for ye all to keep the ball rolling during my absense. Loved the fact that I managed to be pretty much fully disconnected during my time off <3
14:21:50 <alistarle> Hi everyone, following the discussion we had two weeks ago about https://review.opendev.org/#/c/669201/, we just updated it to a full spec :)
14:22:18 <alistarle> Hope it is ready now and complete enough to be reviewed :)
14:22:26 <jokke_> Second, I'd love to get some help with the cluster awareness patch. I think the test flakyness was due to the 60s timeout in there (will look into making that configurable but I reduced it to 20s for now). Any testing and feedback would be greatly appreciated
14:22:55 <abhishekk> jokke_, ack
14:25:42 <jokke_> alistarle: hey, thanks for the spec. Have you guys already looked into writing the plugin for that? I have couple of things in my mind about it.
14:27:12 <jokke_> First: The alternative option, I've been looking into/thinking of making import method that would accept to be called when image is "active" and would be copy between 2 stores (I want to get the cluster awareness done first so we can identify node that has access to both of the stores before doing it)
14:28:01 <abhishekk> Sorry, if we don't have much to discuss, I will like to disconnect little early
14:28:12 <jokke_> Second: I'm all up for allowing import to multiple stores, but I'd way rather prefer using the json for defining that than headers.
14:28:21 <zhengMa> Hi team, I'm expecting more comments on spec https://review.opendev.org/#/c/670454/, thank you!
14:29:10 <alistarle> We proposed the headers because it is already available
14:30:26 <alistarle> We thought that being able to send a list instead of a single store was a smallest change than adding a json
14:30:51 <jokke_> alistarle: the import call uses already the json for it's inputs. For glance-direct it's only the method, for web-download it's the method and uri, same with the planned swift plugin the source was speced into the json payload of the call
14:30:51 <alistarle> but it's totally possible to do it this way and modify the api accordingly
14:31:22 <jokke_> so it will be just defining the parsing of that json input to pick up the data from there
14:31:50 <alistarle> in this case, you suggest to remove the "X-Image-Meta-Store' header?
14:32:07 <jokke_> but I really would like to keep the inputs in one place instead of mixing the json payload of the call and headers
14:33:28 <jokke_> that would be my preference, but obviously depends what others are thinking
14:33:36 <jokke_> just something to think about
14:33:46 <alistarle> yep no problems
14:34:18 <alistarle> we proposed to use the header just because it's already here
14:34:25 <alistarle> for your first point
14:35:11 <jokke_> it's just really silly that we specced that json input into that api to provide all the info user needs to pass to the import call, and then we start mixing headers into the picture and having mixture of the two. Makes it nightmare for client maintainers
14:35:38 <alistarle> yep it totally makes sense
14:36:05 <alistarle> And for your first point, if we allow to import image with 'Active' status
14:36:33 <alistarle> The import could take more time as we should first download from the backend where the data are already available
14:36:48 <alistarle> before uploading to a new one
14:37:24 <jokke_> alistarle: no, so the import method that would be allowed for active image would be just for copying the image data between stores
14:37:32 <jokke_> not providing new data to the image
14:38:09 <abhishekk> what happens if image upload fails to one of the store, is there provision to retry the operation?
14:38:39 <jokke_> so if someone adds new site with new local store, we would allow running "import-store" method providing destination store and glance would copy the data from exisiting store to the destination and update the locations accordingly
14:39:27 <jokke_> abhishekk: I think that's the biggest question around import to multiple stores, should we fail the import if any of the stores requested fails or only if all of them fails
14:39:42 <alistarle> My point was just that to copy the data, glance would have to download it first from store1
14:39:52 <jokke_> alistarle: yes
14:40:01 <alistarle> but in case of a new backend it could be a great feature
14:40:23 <jokke_> alistarle: what I'm saying is not really being either or, I'd be happy with both exisiting
14:40:41 <jokke_> so not really alternative
14:40:49 <abhishekk> IMO we should fail if any of the store fails and revert the whole operation
14:40:53 <alistarle> ok great
14:41:12 <alistarle> If one fails, i think that we should revert the operation
14:41:50 <alistarle> That's what we wrote on the spec
14:41:59 <jokke_> abhishekk: that's my first feeling as well, it's just wasting a lots of resources and time potentially.
14:42:02 <davee_> I can see a use case for having a flag to determine the rollback especially for large images to keep them and resubmit to just the failed stores
14:42:03 <alistarle> and in the linked patch
14:42:31 <abhishekk> jokke_, right
14:42:32 <jokke_> davee_: that's what I'm thinking as well, we could have that on the import input and let user make the call
14:43:44 <jokke_> we could default to fail&revert if any fails and have optional field in the json, something like "fail_if_all_stores_fails": "True"
14:44:12 <jokke_> and in that case the new method of copying the data between stores would be the way to try again
14:44:15 <davee_> jokke: we are on the same page
14:44:25 <abhishekk> I haven't gone through spec yet, but to be clear, if there are 3 stores then we are creating 3 images or just one image with 3 locations?
14:44:40 <alistarle> one image with 3 locations
14:44:40 <jokke_> abhishekk: 1 image, 3 locations
14:45:12 <abhishekk> cool, then we need to test scrubber as well :D
14:46:18 <jokke_> abhishekk: yeah, there is quite a few things to consider in that
14:46:50 <jokke_> lets put our thoughts into that spec review and continue from there
14:46:57 <abhishekk> yes
14:47:28 <alistarle> do you want that we update the specs with what we discussed or wait for the review comments?
14:47:33 <jokke_> anything else, zhengMa I did open your review so it will be on my list of things to look through
14:47:44 <abhishekk> nothing from me
14:47:59 <zhengMa> jokke: Thank you
14:48:22 <jokke_> alistarle: give it like until early next week, I'll put my thoughts into the comments, Everyone, please give your input so alistarle can work new revision next week :)
14:48:24 <zhengMa> nothing for me now
14:48:48 <abhishekk> ack
14:49:03 <jokke_> alistarle: that way you don't need to go back and forth with 25 revisions on that review
14:49:05 <alistarle> ok thanks :)
14:49:21 <jokke_> lets do some brainstorming first :D
14:49:39 <jokke_> ok, if that's all, we get 10min back
14:49:49 <jokke_> going 1st
14:49:58 <abhishekk> Thank you ALL!!!
14:50:09 <jokke_> going 2nd
14:50:27 <jokke_> Sold, Thanks everyone!
14:50:34 <rosmaita> hey, abhishek, need you to look at https://review.opendev.org/#/c/674697/
14:50:34 <jokke_> #endmeeting