14:00:55 <jokke_> #startmeeting glance
14:00:56 <openstack> Meeting started Thu May 10 14:00:55 2018 UTC and is due to finish in 60 minutes.  The chair is jokke_. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:57 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:59 <openstack> The meeting name has been set to 'glance'
14:01:00 <jokke_> #topic roll-call
14:01:05 <jokke_> o/
14:01:06 <smcginnis> o/
14:01:13 <McClymontS> GM/Afternoon all o/
14:01:27 <rosmaita> o/
14:02:19 <jokke_> lets wait for a min if Abhishek is around, he had topic there for the meeting
14:03:00 <rosmaita> smcginnis: while you are waiting, please have a look at https://review.openstack.org/#/c/540089/
14:03:03 <rosmaita> :)
14:03:10 <abhishekk> o/
14:03:17 <jokke_> abhishekk: \o
14:03:29 <jokke_> ok lets get started, looks like we have crew together
14:03:35 <abhishekk> jokke_, o/, I have removed that :D
14:03:41 <smcginnis> rosmaita: On it. ;)
14:03:44 <jokke_> #topic updates
14:04:18 <jokke_> So summit is closing fast, which means that Milestone 2 and with that spec freeze are closing fast too
14:04:27 <jokke_> less than month to go
14:05:29 <jokke_> lets get them reviews in and hopefully our gate is now after the latest test fixes stable enough that we can start getting those features worked and merged we have on our plate
14:06:25 <abhishekk> thanks to brian for that
14:06:31 <jokke_> indeed
14:06:35 * rosmaita blushes
14:06:57 <jokke_> so please prioritize your time to get the features worked unless something critical pops up
14:07:21 <jokke_> I do not want to have majority of the stuff coming in on the FF week, which is not that far away anymore
14:08:51 <jokke_> Then another thing that mugsie kindly flagged, there was discussion around the PY3 migration that community would not need to support the eventlet model of deployments with PY3 ... well we're at least in need of that so lets keep eye on the discussion and chip in
14:09:18 <abhishekk> ok
14:09:45 <jokke_> rosmaita: do we have any updates around releases?
14:10:03 <rosmaita> no, basically said it all last week
14:10:10 <jokke_> yeah, thought so
14:10:27 <jokke_> #topic Glance functional test changes
14:10:33 <jokke_> rosmaita: stage is yours
14:10:52 <rosmaita> thanks ... just want to point out some changes in how the tests are run
14:11:06 <rosmaita> #link https://github.com/openstack/glance/blob/master/tox.ini#L38-L53
14:11:18 <rosmaita> that's the relevant part of our current tox.ini in glance
14:12:05 <rosmaita> so what we're doing now is first running the "regular" functional tests
14:12:16 <rosmaita> and then running the ones that must be run sequentially
14:12:29 <rosmaita> so, the output will look slightly different
14:12:44 <rosmaita> here's an example
14:12:55 <rosmaita> #link http://logs.openstack.org/79/566879/2/check/openstack-tox-functional/a6bee54/
14:13:21 <rosmaita> first, if you open that link, click on "testr_results.html.gz"
14:13:47 <rosmaita> all the test results, both concurrent and serial, are combined into that report
14:14:09 <rosmaita> (look for glance.tests.functional.serial aobut halfway down the page)
14:14:19 <rosmaita> so that's cool
14:14:36 <rosmaita> if you look at the job-output.txt.gz it's a bit different
14:14:49 <rosmaita> if you scroll up from the bottom, you'll see the 9 serial tests passing
14:15:00 <rosmaita> you have to keep scrolling up to get the failure from the concurrent test
14:15:01 <abhishekk> yes, two sets of results
14:15:09 <rosmaita> right
14:15:27 <rosmaita> and that's what you should see running locally
14:15:38 <rosmaita> so that's the only difference, really
14:15:44 <rosmaita> just want everyone to be aware
14:16:15 <rosmaita> i think that's it ... any questions?
14:16:57 <jokke_> So the DL;DR is that you might not find the failure anymore at the bottom of the job-output due to the fact that the functinal tests are split to 2 different runs
14:17:01 <abhishekk> just one, instead of having a file can we use regex in the tox?
14:17:15 <rosmaita> abhishekk i don't think so
14:17:21 <rosmaita> there's a blacklist-regex option
14:17:25 <rosmaita> but not for the whitelist
14:17:39 <rosmaita> not sure why
14:18:03 <rosmaita> i think the file is the preferred method
14:18:05 <abhishekk> because in future if  we want to add more sequential tests then we need to add another file and it will also increase the result set
14:18:05 <jokke_> tbh, I'd prefer not ... regexp is really by no means human readable ... characters are pretty cheap ;)
14:18:09 <rosmaita> the blacklist-regex is mainly for testing your regex
14:18:09 <smcginnis> Looks like a good approach for this. We can be explicit by editing the serial-functests.txt file.
14:18:24 <rosmaita> abhishekk won't need a new file
14:18:27 <abhishekk> or we can add in the same file
14:18:29 <jokke_> abhishekk: no, we just add those to the serial tests file
14:18:33 <rosmaita> just one regexp per line
14:19:00 <abhishekk> rosmaita, got it, I was reading that file as scrubber-functests.txt :(
14:19:06 <rosmaita> ah, ok
14:19:13 <rosmaita> i tried to be proactive in naming :)
14:19:23 <jokke_> :)
14:19:30 <abhishekk> :D
14:19:49 <rosmaita> btw, i'm not sure all those scrubber tests need to be run sequentially
14:20:03 <rosmaita> but i didn't want to take the time to figure out which ones
14:20:21 <jokke_> that's fine
14:20:31 <rosmaita> refactoring opportunity for anyone with too much time on their hands
14:20:37 <jokke_> the situation is better than it was and we have logical gate atm.
14:20:43 <rosmaita> agreed
14:20:46 <abhishekk> +2
14:21:42 <jokke_> ok, anything else about this topic?
14:22:03 <rosmaita> sort of related, i have one more patch up related to this
14:22:27 <rosmaita> (sorry couldn't find link)
14:22:33 <rosmaita> #link https://review.openstack.org/#/c/566947/
14:23:22 <abhishekk> yeah, that is a good change
14:23:24 <rosmaita> i have a gut feeling that some of the subunit parser problem was from hammering the same test with a bunch of the same assertions, but i have no data
14:23:33 <rosmaita> anyway, that change simplifies things a bit
14:24:03 <rosmaita> but it will also only allow retries for one paraticular proiblem, not any assertion failure
14:24:07 <jokke_> I saw that in the list but honestly haven't even looked at it yet
14:24:22 <rosmaita> ok, just so you're aware of it
14:24:26 <jokke_> thnx
14:24:32 <rosmaita> abhishekk and wxy like it!
14:25:01 <rosmaita> ok, that's it from me on this topic
14:25:12 <jokke_> Any more questions anyone?
14:25:31 <abhishekk> no
14:25:36 <jokke_> going once
14:25:43 <smcginnis> No, looks good on first quick pass.
14:25:45 <jokke_> twice
14:25:52 <jokke_> sold
14:25:53 <McClymontS> Im good
14:26:01 <jokke_> #topic Open Discussion
14:26:23 <jokke_> so I think abhishekk wanted to highlight the bug meeting
14:26:27 <jokke_> or was that rosmaita
14:26:31 <rosmaita> that was me
14:26:32 <jokke_> we have plenty of time
14:26:48 <rosmaita> just wanted to remind people, since it's every other week
14:27:27 <rosmaita> so, for anyone reading through the meeting log, bug squad meeting at 1000 utc monday
14:27:37 <rosmaita> #link http://eavesdrop.openstack.org/#Glance_Bug_Squad_Meeting
14:28:03 <abhishekk> rosmaita, we can change time if it's too early for you
14:28:22 <rosmaita> no, it works for me
14:28:23 <McClymontS> I am also EDT
14:28:35 <McClymontS> rosmaita is stronger than I
14:28:43 <rosmaita> :)
14:28:46 <abhishekk> great :D
14:28:47 <McClymontS> I'll be rooting for you guys from my slumber
14:29:19 <rosmaita> what i think we should do is let people tell us if the time is no good
14:29:28 <rosmaita> and then we can re-negotiate
14:29:36 <abhishekk> sounds good
14:29:45 <abhishekk> rosmaita, I have found TODO on your name, https://github.com/openstack/glance/blob/master/glance/tests/functional/__init__.py#L79
14:29:48 <rosmaita> but i think since the weekly glance meeting is not so convenient for APAC people, this is a good compromise
14:29:59 <abhishekk> do we intend to change that??
14:30:03 <jokke_> sounds good. abhishekk do you want to talk about the 'glance-direct' situation to have clarification for that? I know I was not even shortworded on the bug and patch as I saw the topic in he meeting agenda
14:30:34 <abhishekk> jokke_, I already know that it gonna be rejected :D
14:31:02 <rosmaita> jokke_ when is emable_image_import being removed?
14:31:03 <abhishekk> but for the other peoples to know, I had reported a bug https://bugs.launchpad.net/glance/+bug/1770325
14:31:04 <openstack> Launchpad bug 1770325 in Glance "Remove glance-direct from enabled_import_methods config option" [Undecided,Invalid] - Assigned to Abhishek Kekane (abhishek-kekane)
14:32:15 <abhishekk> What I was thinking is that most of the glance setup will have HA enabled and as glance-direct is allowed it might be problematic to use that without shared storage
14:32:37 <jokke_> rosmaita: I've been holding off the patch for summit forum sessions, wanna hear from operators before rather than after merging if they are really really against it
14:32:47 <McClymontS> Good call jokke_
14:33:19 <rosmaita> abhishekk i guess the shared storage isn't a problem for web-download because it's being handled by a single task?
14:33:31 <jokke_> rosmaita: that's correct
14:33:35 <abhishekk> rosmaita, yes
14:33:44 <jokke_> it's rather single request so same node handles it every time
14:33:54 <rosmaita> ok
14:33:57 <abhishekk> should document it somewhere?
14:34:05 <jokke_> I think it is
14:34:10 <abhishekk> s/should/should we?
14:34:27 <rosmaita> we should probably review the doc to see
14:34:52 <jokke_> not sure if it was in the config help texts, release notes, admin guide, os some of the other about dozen logical places
14:35:00 <abhishekk> then no worries, but I haven't found that, may be have a look once again
14:35:04 <jokke_> s/os/or/
14:35:20 <rosmaita> well, if we have extra time today, http://git.openstack.org/cgit/openstack/glance/tree/doc/source/admin/interoperable-image-import.rst
14:35:38 <rosmaita> i think maybe it's in release notes, maybe not the admin doc
14:35:50 <rosmaita> and we know no one reads release notes :P
14:36:01 <abhishekk> :D
14:36:20 <rosmaita> actually, see line 63-65 in that doc
14:36:28 <jokke_> rosmaita: only thing that is our problem about it, is that we need to keep closing those invalid bugs ;P
14:36:32 <rosmaita> maybe we could make them bold or italic
14:36:41 <jokke_> no-one reads the other docs either :D
14:36:51 <rosmaita> :)
14:37:00 <McClymontS> Ever
14:37:25 <abhishekk> it is documented :D
14:37:45 <jokke_> like was proven when this request came up as TripleO had not handled the fact that those methods are enabled by default and the shared FS is needed for HA ;P
14:38:27 <abhishekk> :D
14:38:54 <rosmaita> oops
14:39:09 <jokke_> so that gives us a bridge for my next result of brainstorming
14:39:23 <rosmaita> maybe we need new glance t-shirts: Glance -- READ THE RELEASE NOTES
14:39:32 <jokke_> should we finally consider making glance cluster aware?
14:39:39 <McClymontS> rosmaita that was in my email signature for a while
14:39:45 <McClymontS> minus the Glance -- part
14:39:46 <rosmaita> :)
14:40:15 <rosmaita> jokke_ how do you mean?
14:40:16 <abhishekk> yes we need t-shirts ;P
14:41:22 <jokke_> what I mean is, if we changed the import call handling to be message sent to message queue with the request serialized as payload we could get the node picking the import up that has the data, rather than forcing everyone going for shared filesystem and all the quirkyness of that
14:41:23 <smcginnis> jokke_: Are you thinking something like this: https://specs.openstack.org/openstack/cinder-specs/specs/mitaka/cinder-volume-active-active-support.html
14:41:55 <smcginnis> jokke_: A little different I guess.
14:42:04 <jokke_> I think we have loads of things in glance that could be improved with such
14:42:06 <rosmaita> on previous topic, also see lines 611-612 in http://git.openstack.org/cgit/openstack/glance/tree/etc/glance-api.conf
14:42:16 <jokke_> like image-cache
14:42:30 <jokke_> when we could actually provide usable precaching
14:44:12 <jokke_> it would not be difficult to implement and we could move stuff that really benefits of it gradually when those things pops up
14:44:45 <abhishekk> jokke_, i guess we should have that for S
14:44:46 <jokke_> that would also provide us a way to finally provide actual worker nodes for the tasks if we wanted to
14:44:56 <McClymontS> I think next cycle makes sense probably right?
14:44:58 <abhishekk> ++
14:45:03 <jokke_> definitely not for Rocky
14:45:10 <McClymontS> Also a huge +1 on worker nodes
14:45:28 <jokke_> this is just idea I wanted to probe your feelings about now when you're all here
14:46:47 <jokke_> if you think it's something we could consider, I could roll it vial operators in the summit and start writing the spec for next cycle
14:47:00 <abhishekk> we need another service i guess which will run on all the nodes?
14:47:23 <jokke_> abhishekk: nope, we can have the API listening the queue
14:47:42 <abhishekk> jokke_, ok
14:47:45 <jokke_> I rather not have yet another service just when we're getting rid of registry
14:47:53 <rosmaita> ++
14:47:56 <abhishekk> :D
14:48:23 <jokke_> message queues are brilliant for this kind of communication
14:48:40 <abhishekk> agree
14:49:02 <jokke_> and for the worker nodes, they can be for example fenced off with firewall just having access to the message queue, database and storage
14:49:15 <jokke_> doesn't matter if they ever receive a single api call
14:49:55 <McClymontS> I like that a lot from the security perspective too jokke_
14:49:55 <jokke_> so we would not need to refactor the whole shebang to make this happen
14:50:12 <rosmaita> there seem to be a lot of pluses to this idea
14:50:23 <abhishekk> exactly
14:51:01 <McClymontS> I'd definitely like to be involved in reviewing/writing any parts of that spec that make sense
14:51:23 <jokke_> cool, I'll add it to my list for the summit and start looking into this
14:52:04 <jokke_> as I haven't seen yet anyone screaming "NOOOO" unless smcginnis is banging head against the wall instead of typing :P
14:52:19 <abhishekk> lol
14:52:24 <smcginnis> jokke_: Hah, no, it sounds like a good plan to me. :)
14:52:43 <abhishekk> he might be thinking of implementation :D
14:52:53 <jokke_> ok, expect hearing more during/after the summit then! :)
14:53:31 <rosmaita> we can whiteboard it there and do some screaming
14:53:37 <jokke_> yup
14:53:41 <abhishekk> :D
14:53:41 <McClymontS> Awesome
14:53:58 <jokke_> ok, 7min anything else?
14:54:47 <abhishekk> nope, others please have a look on multi-store specs (https://review.openstack.org/#/c/562467/)
14:54:49 <rosmaita> got a dumb patch up for glance_store docs if anyone wants to look: https://review.openstack.org/#/c/565830/
14:54:59 <rosmaita> abhishekk left a comment there for you
14:55:05 <McClymontS> I am good to go, will take a look
14:55:14 <jokke_> ohh
14:55:19 <abhishekk> rosmaita, thank you, added a reply inline
14:55:33 <jokke_> McClymontS: the multihash moving at all?
14:55:34 <rosmaita> cool, i will finish reading through the api this afternoon
14:55:52 <rosmaita> i think the multihash is moving at the speed of rosmaita
14:55:58 <McClymontS> I know rosmaita had said he was looking at some stuff
14:56:15 <McClymontS> My time has been very limited lately unfortunately
14:56:20 <rosmaita> now that the functional tests are stable for a while, i will concentrate on multihash
14:56:25 <abhishekk> rosmaita, McClymontS thank you
14:56:29 <rosmaita> need to get McClymontS some feedback
14:57:00 <abhishekk> rosmaita, sure thing
14:57:05 <jokke_> rosmaita: McClymontS: good to know it's still in minds. Lets try to give it some love finally
14:57:10 <rosmaita> ok
14:57:23 <rosmaita> abhishekk to answer your question a while back about the TODO ...
14:57:47 <rosmaita> it needs to wait until jokke_ gives the ok to remove the enable_image_import option
14:57:54 <abhishekk> (between I almost have a code ready, except jokke_ wants new way for download)
14:58:03 <rosmaita> but i guess saying Queens there was a bit premature
14:58:06 <abhishekk> rosmaita, ok
14:58:21 <smcginnis> How are we on the series goals? I think both are either done or don't affect us? Not sure on all repos though.
14:58:23 <jokke_> rosmaita: not the first gun we've jumped on :D
14:58:36 <abhishekk> rosmaita, exactly, or we should change it to True?
14:58:58 <rosmaita> i think it did get changed to True
14:59:03 <jokke_> smcginnis: correct, we are good
14:59:11 <smcginnis> jokke_: Excellent!
14:59:26 <rosmaita> abhishekk: http://git.openstack.org/cgit/openstack/glance/tree/etc/glance-api.conf#n645
14:59:36 <jokke_> ok, time; Thanks all!
14:59:38 <abhishekk> rosmaita, it's still false in the tests
14:59:40 <smcginnis> Thanks!
14:59:41 <abhishekk> thank you all
14:59:45 <jokke_> #endmeeting