Tuesday, 2015-07-28

openstackgerritSteve McLellan proposed openstack/searchlight: Add 'sort' parameter  https://review.openstack.org/20626800:17
*** openstackgerrit has quit IRC00:46
*** openstackgerrit has joined #openstack-searchlight00:47
*** sigmavirus24 has quit IRC02:07
*** lakshmiS has joined #openstack-searchlight06:46
TravTkrykowski: if you have a chance, can you look at these patches: https://review.openstack.org/206242  & https://review.openstack.org/#/c/202392/11:24
*** lakshmiS has quit IRC13:48
*** sigmavirus24_awa has joined #openstack-searchlight13:57
*** sigmavirus24_awa is now known as sigmavirus2413:57
TravTsigmavirus24: would you be able to look at this one again? https://review.openstack.org/#/c/206242/16:04
sigmavirus24TravT: took a quick look but I don't have time to do a deep dive on it so I left a Q16:07
TravTOk, thanks!16:08
sigmavirus24Depending on the answer to the Q I can +2 it or -1 =P16:09
TravTwell, we'll see what sjmc7 says.  The spot you put the question on is what is called during a regular indexing call (not via notifications)16:13
TravTeg searchlight-manage index sync call16:13
sjmc7my ears are burning16:14
sjmc7good catch sigmavirus24, thought i'd removed that16:14
sigmavirus24ah16:14
sjmc7one sec, will push up a patch directly16:14
sigmavirus24TravT: see I thought it might be something like that16:14
sjmc7:)16:15
sigmavirus24But I wasn't sure because everything else was changes to use the "notification" instead of "image" function16:15
sigmavirus24so16:15
* sigmavirus24 shrugs16:15
* sigmavirus24 just asks questions that are sometimes insightful16:15
TravTit is always good to ask questions... but now I'm wondering about the flow order.16:15
sigmavirus24heh16:16
* sigmavirus24 sits backs and relaxes all cool16:16
TravTmight have to step through it in the debugger again, because i would have thought it was line 81 that would go away not line 6316:16
openstackgerritSteve McLellan proposed openstack/searchlight: Glance image notification serialization fix  https://review.openstack.org/20624216:17
sjmc7well, now's the time to ask. removed those lines16:18
sigmavirus24sjmc7: https://github.com/openstack/searchlight/blob/55471cb2f18a7c54bf0ee627191d58bf23ca33b5/searchlight/elasticsearch/plugins/glance/images.py#L12316:19
sigmavirus24that might need the is_public stuff16:19
TravTThanks, sjmc7, I'll run it through its paces... I am going to run debug through both notification flow and index flow... but I think I see why I was confused on that.16:19
* sigmavirus24 did a quick github search since TravT brought it up16:19
sjmc7the API returns 'visibility'16:19
sigmavirus24Ah16:19
sjmc7although... i wonder about v1 images16:19
sigmavirus24We're only using v2 images API?16:19
sigmavirus24v1 uses is_public16:19
sigmavirus24Whatever we do, we should leave a comment explaining it16:19
sigmavirus24=P16:20
sigmavirus24or make a separate util function so we aren't copying a ternary around16:20
* sigmavirus24 hides again to figure out this keystone v3/swift glance store thing16:20
TravTthat sounds like loads of fun16:20
TravTBTW, sjmc7 as we start looking at this from horizon patches, I have a withering patch I was working on with Julie Pichon to support transition horizon from v1 to v2 glance.16:26
TravThttps://review.openstack.org/#/c/150084/4/openstack_dashboard/api/glance.py16:26
sjmc7sorry, modem reboot16:26
sjmc7v1 <-> v2 compat is the subject of another patch i think16:27
TravTbut I do believe what you have done means v1 won't work for indexing.16:27
sjmc7i doubt this is the only thing16:27
sjmc7i haven't changed anything for v1 :)16:27
TravTno you did it16:27
sjmc7if v1's broken, it's broken16:27
sjmc7i did many things but not that16:27
* TravT just wanted to blame sjmc716:27
sjmc7this osunds like a Meatloaf song16:27
TravTok, i have about 40 minutes. are you thinking you'll do more with that patch or not?  if so, i'll spend my time on other things.  if so, I will run it through its paces again.16:29
TravT2nd if so should be if not16:29
sjmc7so yes, it sounds like v1 may be unhappy. but i don't think that's made better or worse by this patch - this *should* only affect notifications which apparently are always v1. well, the only thing is whether i change it to always call the API16:29
TravTyou know16:29
sjmc7as we briefly discussed yesterday. and maybe that's better16:29
TravTi do think in the long run that will be better.16:30
sjmc7ok. i can certainly do that16:30
sjmc7in which case, do something else for your 40 minutes16:30
TravTbut, we could make that an item of discussion on thursday's meeting when we have more attention16:31
TravTi'd like to get a working glance listener16:31
sjmc7yeah. maybe let's get this in and then can rework later16:31
sjmc7since right now it's broken out of the box16:31
TravTi would not complain if the visibility / public was its own function16:31
TravTwhether or not the regular serialize function calls it isn't totally necessary at this point16:32
sjmc7if i do that i might as well call it there too16:32
sjmc7i can make that change leter16:32
TravTok, well, i'll put this patchset through its paces.16:35
sjmc7gimme two secs and i'll put that change up16:35
TravTok.16:35
TravTI'll be back in a couple minutes myself then.16:35
openstackgerritSteve McLellan proposed openstack/searchlight: Glance image notification serialization fix  https://review.openstack.org/20624216:56
sjmc7sorry TravT, my turn for a bad ISP16:57
sjmc7patch is up16:58
TravTsjmc7, just looking now16:59
TravTi like that its pulled out.16:59
TravTi guess no discrete unit test since it is _function?16:59
sjmc7yeah16:59
TravTokay, i was just going to start hacking at another patch, but I'll spend the next ten minutes running this instead and see where I get.17:00
sjmc7the notification unit test (and any v1 format test we don't have) should cover it17:00
TravTalthough there is not expected of "visibility" : "private"17:01
sjmc7i can put another test in17:01
TravToh wait, maybe there is in the other tests in that file?17:01
sjmc7there are some private tests, yeah. thiough they start with a v2-like input17:02
TravTprobably better to go ahead and add a notification test where is_public starts false.17:02
sjmc7ok dokey. give it a test as-is, i won't change anything functionally17:03
TravTi'm getting pinged in hipchat for stuff now... :S17:04
sjmc7:)17:04
TravTthis is what i get for logging back into hipchat17:06
TravTsjmc7, i'll be out for the next hour or so. but will go through validating that patch as soon as I'm back.17:14
sjmc7sure thing17:14
sigmavirus24sjmc7: TravT do you know lakshmi?18:04
sjmc7yep18:04
sigmavirus24Could you have him look at https://review.openstack.org/195775 once more?18:04
sigmavirus24Eric responded, and I think we're good to go, but lakshmi and krykowski are SMEs18:05
sigmavirus24oh krykowski could you look at https://review.openstack.org/195775 too?18:05
sjmc7mailed him, but he's in India so likely won't be til tomorrow18:06
sigmavirus24sjmc7: no worries18:19
sigmavirus24Not a rush18:19
sigmavirus24It just has a +2 and I wanted to make sure his concern(s) were addressed by the reply18:19
sjmc7very courteous!18:20
openstackgerritSteve McLellan proposed openstack/searchlight: Glance image notification serialization fix  https://review.openstack.org/20624220:50
TravTsjmc7: ^ does that mean I should test out the patch now?20:51
TravTanything more to come?20:51
sjmc7have you been waiting all afternoon??20:51
sjmc7no, i'm done20:51
sjmc7it was a good suggestion20:51
TravTok.  I wasn't explicitly waiting for your patch.  I got pulled into several other things.20:52
sjmc7just yankin your chain. i think it's done, and i also think on reflection it would be better to switch to use the API for now, at least until notifications give us everything20:52
TravTi still think notification payload will be a problem in general20:54
TravTbut, let's look through this one.20:54
TravTbtw, i found a fun way to confuse yourself with notifications earlier20:55
TravTset up glance to publish to notifications, searchlight_indexer topic20:56
TravTthen have searchlight_listener service running20:56
TravTand start up searchlight listener in pycharms debugger20:56
TravTthen watch your hair fall to the ground as you try to figure out why your debugger breakpoints  don't seem to affect the index as you cross reference via elastic search direct queries20:57
sigmavirus24TravT: I don't even understand that21:03
sigmavirus24then again I use vim21:04
sigmavirus24so21:04
sigmavirus24I've already lost all my hair21:04
sigmavirus24If you want to pull your hair out21:04
sigmavirus24Use Glance + Keystone v3 + Swift and watch as nothing gets logged from glance_store about the auth failure and you get an error that makes it seem like nothing even reached glance_store yet from glance v1 api21:05
TravTooh, that seems like a great way to lose your hair.21:05
sigmavirus24Oh yeah21:09
sigmavirus24Then I remembered Jamie Lennox had a patch fixing it21:09
sigmavirus24Then I didn't look closely enough and it still didn't work21:09
sigmavirus24Then i read his patch very closely21:10
sigmavirus24And saw that there are new config options I have to set21:10
sigmavirus24And then I got it working21:10
TravTlovely21:14
TravTI hope you requested better documentation.21:15
TravTsigmavirus24:  I just locally reverified the serialization fix. When you are able to look again, would appreciate it.  https://review.openstack.org/#/c/206242/21:57
TravTsjmc7 ^21:57
sigmavirus24One comment in-line but it's not really important for that patch to be approved (so I approved it)22:01
openstackgerritMerged openstack/searchlight: Glance image notification serialization fix  https://review.openstack.org/20624222:06
sigmavirus24fast gate is fast22:07
TravTwow22:07
TravTi walked away to get a drink of water, come back and it merged.22:08
* sigmavirus24 has many things in the air at this moment22:08
sigmavirus24I left the actual testing to you TravT22:08
TravTgood comment22:08
sigmavirus24I just looked at the code22:08
sigmavirus24I trust you tested it thoroughly22:08
TravTtesting it has consumed a good portion of my day.22:08
sigmavirus24I've been making glance + swift + keystone v3 work for most of the day22:08
sigmavirus24when I wasn't participating remotely in glance midcycle discussions22:09
sigmavirus24or digging into cryptography22:09
sigmavirus24or ... crying22:09
TravT:P22:09
TravTi found a few error handling items while testing this22:09
TravTnot related to this patch directly22:10
TravTthat I might put up a patch to start a conversation about.22:10
openstackgerritTravis Tripp proposed openstack/searchlight: Change Glance notifications to be create_or_update  https://review.openstack.org/20675622:52
*** sigmavirus24 is now known as sigmavirus24_awa23:07
openstackgerritTravis Tripp proposed openstack/searchlight: Glance index visibility should only be for v1  https://review.openstack.org/20676223:35

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!