14:01:13 <tellesnobrega> #startmeeting sahara 14:01:13 <openstack> Meeting started Thu Jul 12 14:01:13 2018 UTC and is due to finish in 60 minutes. The chair is tellesnobrega. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:01:14 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:01:16 <openstack> The meeting name has been set to 'sahara' 14:01:53 <tellesnobrega> show off hands o/ 14:02:02 <jeremyfreudberg> o/ 14:02:04 <tosky> o/ 14:02:06 <zchkun> o/ 14:02:41 <tellesnobrega> #topic News/Updates 14:03:28 <tellesnobrega> I'm currently testing the ssl problem in ambari 14:03:39 <tellesnobrega> HDP 2.3 with ambari 2.2.0.0 works 14:03:48 <jeremyfreudberg> i am hard at work on s3 and apiv2 ecosystem. rushing to finish everything 14:03:51 <zchkun> I'm working on MapR6 upgrade. The current phase is creating the image with sahara-image-pack. 14:04:06 <tellesnobrega> next in line is HDP 2.4 with ambari 2.2.1.0 14:04:15 <tosky> jeremyfreudberg: I have few questions about S3, but for later 14:04:24 <jeremyfreudberg> tosky: certainly 14:04:34 <tellesnobrega> we can have a topic on it 14:04:37 <tellesnobrega> and apiv2 for sure 14:04:39 <tosky> I helped testing some of the patches for new plugins and S3 and gently nagged people to merge my patches :) 14:04:56 <tellesnobrega> other than that I'm also working on spark on sahara-image-pack 14:06:50 <tellesnobrega> any more news? 14:07:15 <jeremyfreudberg> just a reminder that i am gone the next two weeks.... or rather, unable to write code ( i can still review ) 14:07:43 <tellesnobrega> ok 14:08:22 <tellesnobrega> lets move on 14:08:28 <tellesnobrega> #topic S3 14:08:33 <tellesnobrega> tosky, you have the floor 14:11:14 <tosky> I merged the S3 patch, but I'm not 100% sure about the differences between using s3:// and no http/https vs s3a:// and the need for it 14:11:49 <tosky> the definition of S3 resources for job binaries and data sources if different right now 14:12:11 <jeremyfreudberg> yes, the definition is different, and intentionally 14:12:23 <jeremyfreudberg> one is designed to work with boto, and the other is designed to work with hadoop-aws.jar 14:13:33 <tosky> but we are abstractor (!) 14:14:13 <tosky> can't we provide a unified way for users? 14:15:13 <jeremyfreudberg> maybe 14:15:19 <tellesnobrega> that would be the best way 14:15:33 <jeremyfreudberg> they aren't exactly the same thing, though 14:15:43 <tellesnobrega> we aim to be a front, makes things easier on the client side 14:15:50 <jeremyfreudberg> a job binary is a single file, but a data souce may be a directory or a wildcard 14:16:17 <jeremyfreudberg> i do see your point, though 14:16:18 <tosky> but isn't it something that we can catch on validation, before using the data source or the job binary? 14:16:46 <jeremyfreudberg> i can definitely add validation for the http or not thing 14:16:51 <jeremyfreudberg> there is already validation for s3 / s3a 14:17:22 <tosky> that would be better, and also the s3 vs s3a 14:17:35 <tosky> if we can hide the difference, that would be good too 14:18:52 <jeremyfreudberg> i am not sure i want to create a new unified addressing format 14:18:56 <jeremyfreudberg> that may be confusing for users too 14:19:58 <tellesnobrega> how is it used outside of sahara? 14:19:58 <tosky> do you see a use case where users may end up using the address (for data sources, if we unify to s3:/) wrongly 14:20:00 <tosky> ? 14:20:41 <tosky> or the other way: if it's easier for people to use this on hadoop stuff, maybe we could go with s3a also for job binaries 14:20:49 <jeremyfreudberg> yes, because for example, spark jobs take everything in as a positional arguemnt, either datasource://<ds-uuid> or s3a://actual/path 14:21:15 <jeremyfreudberg> it would be wrong to do s3a for job binaries, because they aren't s3a 14:21:57 <tosky> but we know that, do users need to do that? 14:22:19 <tosky> I know that they are not s3a; I'm just saying to do the conversion on the fly before using them 14:22:32 <tosky> before passing them to the code that uses them 14:23:17 <jeremyfreudberg> hmm 14:23:27 <tosky> same for the other way: in case of using s3:/, that would need to be changed to s3a:/ when passed to the spark job or other users 14:24:02 <tosky> the idea is to hide the different protocol on the sahara side; the different code for job binaries and data sources would see the protocol required 14:25:30 <tellesnobrega> I like this approach, maintains our UX improvement that we aim every cycle 14:25:56 <jeremyfreudberg> i will have to think about it 14:27:00 <tellesnobrega> jeremyfreudberg, give some thought and let us know how feasible it is, if it is too complicated, with all the apiv2 stuff we need fixing, we can postpone it, but it would be nice to have it 14:27:34 <tosky> uhm, we can't postpone it too much, because then we would have to support both schema 14:27:51 <jeremyfreudberg> yep 14:27:54 <tosky> but that wouldn't be a problem I guess, we can just not document the old non-unified option 14:27:55 <tellesnobrega> tosky, true 14:27:55 <tosky> anyway 14:28:19 <jeremyfreudberg> s3 is a big topic - anything else on it? 14:28:34 <tellesnobrega> not from me 14:28:56 <jeremyfreudberg> tosky? 14:29:02 <tosky> nothing else right now 14:29:08 <jeremyfreudberg> oki 14:29:32 <tellesnobrega> lets move on 14:29:33 <tosky> I will have more questions when I do more work on the tests, probably, but for now it's fine 14:29:51 <tellesnobrega> #topic APIv2 14:30:25 <tellesnobrega> jeremyfreudberg, I'm a little outdated on status on apiv2, can you give a quick summary of is done and what is still needed and how far we are from finishing it 14:31:08 <jeremyfreudberg> yep: 14:32:44 <jeremyfreudberg> in the past month, i made changes in APIv2's create multiple clusters and force delete, fixed some other (final )inconsistencies, and am now working on microversion(ish) and dashboard 14:33:28 <jeremyfreudberg> still to do (maybe in Stein) will be OSC for APIv2, good api-ref, and testing 14:35:57 <tosky> can we publish api-ref for v3 right now? I think it should be easy 14:36:07 <tosky> s/v3/v2/ :D 14:36:17 <tellesnobrega> I got worried a bit about v3 14:36:19 <tosky> it would help also as reference for the tempest client 14:36:21 <jeremyfreudberg> tosky -- it's not as magic as we'd like, there is some manual stuff to do for api ref 14:36:43 <tellesnobrega> jeremyfreudberg, I can start working on OSC 14:36:45 <jeremyfreudberg> also, we need to decide if we want to go EXPERIMENTAL->CURRENT in rocky 14:36:49 <tellesnobrega> hopefully I can finish in time 14:37:01 <tellesnobrega> we need OSC to have it current right? 14:37:03 <tosky> CURRENT means enabled, but not default, right? 14:37:12 <jeremyfreudberg> tosky: well, the user requests his version 14:37:22 <jeremyfreudberg> osc and dashboard could still default to 1.1 14:37:37 <jeremyfreudberg> but if we make it CURRENT, then that means all future changes will have to be in microversions 14:37:43 <tosky> and with old unmodified clients, you would still get the old version? 14:38:36 <tellesnobrega> I believe so 14:38:50 <tellesnobrega> I vote we go for CURRENT, but I would like to have OSC done as well 14:39:00 <tellesnobrega> client freeze is next week? 14:39:06 <tosky> I would vote to go for CURRENT only if scenario tests are passing 14:39:17 <tosky> which means support in saharaclient 14:39:22 <tosky> which means probably not enough time 14:39:30 <tellesnobrega> that seems like a good line 14:39:43 <jeremyfreudberg> tellesnobrega: client library freeze is same as rocky-3 14:39:59 <jeremyfreudberg> tosky: support is in saharaclient 14:40:00 <tellesnobrega> july 26th 14:40:02 <tosky> otherwise I don't see it too sound to say "yes, it's stable, but if you want to have it working you need microversining 2.5" 14:40:21 <jeremyfreudberg> i'm happy to do EXPERIMENTAL for one more cycle 14:40:38 <tellesnobrega> I guess I will have time to finish by july 26th 14:41:32 <jeremyfreudberg> now that we will have dashboard (and hopefully OSC), that will serve as its own kind of testing (through use) 14:41:44 <jeremyfreudberg> so we'll have one cycle more to naturally discover problems 14:41:57 <tosky> if you don't mind, unless you think you are working on that or something that touches that, I can take a look at api-ref generation 14:42:01 <tellesnobrega> but I would say, keep experimental for now. but in a couple weeks, if we are more confident we can set the target to CURRENT 14:42:29 <jeremyfreudberg> tosky: please work on api-ref. server side "should" already be done by now 14:42:32 <tosky> oki 14:42:56 <tellesnobrega> awesome 14:42:58 <tellesnobrega> do we have a plan? 14:43:17 <jeremyfreudberg> tellesnobrega: i will propse a patch that brings it to CURRENT, and we probably won't merge it, but it will be there just in case 14:43:23 <tosky> on the other side, switching to CURRENT at the beginning of cycle would allow maybe some fixes without having to change microversioning? 14:43:48 <tosky> or marking it as CURRENT triggers the need for proper microversioned changes even if it's not released yet? 14:44:09 <jeremyfreudberg> tosky: i'm fairly certain a CURRENT api has to be a stable api... but i'm not sure exactly how that works relative to release cycle 14:44:27 <tosky> need to dig 14:44:29 <tosky> ok 14:44:34 <tellesnobrega> I think inside the cycle we can still make changes without microversioning 14:44:36 <tosky> anyway, we have some required blockers 14:44:37 <tellesnobrega> for the first one 14:45:11 <jeremyfreudberg> tosky - is anything blocking, or is all just "very nice to have"? 14:45:24 <jeremyfreudberg> and by very nice, i mean, very very very nice 14:45:48 <tosky> I would say that the first iteration of API v2 should work in the same way as the current v1 from the point of view of the gates 14:45:58 <tosky> hence scenario tests (and probably tempest tests too) 14:46:40 <jeremyfreudberg> tosky: my dashboard patch (at least when it is completed) will give you a good idea of how saharaclient interaction has changed from apiv1.1->apiv2, and you can incorporate that into sahara-tests 14:46:59 <tellesnobrega> by same way you mean 1 to 1 feature wise? 14:47:00 <tosky> oki 14:47:18 <tosky> tellesnobrega: well, we should not have visible regressions 14:47:39 <tosky> of course we have something less (did we cut internal job binaries already?) and some features more 14:47:41 <tellesnobrega> no regressions for sure 14:47:46 <tellesnobrega> can we have more features? 14:47:49 <jeremyfreudberg> tosky: apiv2 has no JBI 14:48:15 <jeremyfreudberg> tosky: APIv2 new features include force delete, and decommission specific node 14:48:15 <tellesnobrega> by that I mean, boot_from_volume 14:48:21 <jeremyfreudberg> tosky: and yes, boot from volume 14:48:53 <tellesnobrega> I will need to work decommission on the dashboard but we have it on apiv2 14:49:07 <tosky> we already have more features: my baseline is that, apart from the changed API and the know removed feature, *at least* the operations available also previously should work 14:49:28 <jeremyfreudberg> tosky: yep 14:49:33 <tellesnobrega> totally agree 14:49:34 <tosky> and by "work" I mean "in a reproducible way on zuul" 14:49:35 <tosky> oki :) 14:49:39 <jeremyfreudberg> tellesnobrega: i am trying to get decomission node in dashboard done this week 14:49:59 <jeremyfreudberg> if i don't finish it, i would love your help. you will have to do boot from volume yourself 14:50:03 <tellesnobrega> jeremyfreudberg, I appreciate that 14:50:19 <tellesnobrega> boot from volume is done 14:50:25 <tellesnobrega> on dashboard as well 14:50:33 <tellesnobrega> it is only a check_box 14:51:05 <jeremyfreudberg> tellesnobrega: oh, right, but you will have to adapt into the apiv2 way of dashobard things, after my patch 14:51:19 <tellesnobrega> jeremyfreudberg, sure 14:51:48 <jeremyfreudberg> tellesnobrega: i also suggest that you split out the saharaclient boot from volume into 2 separate patches 14:51:51 <jeremyfreudberg> basic client, and then osc 14:52:00 <tellesnobrega> ok 14:52:00 <jeremyfreudberg> the basic client must be merged this cycle for dashboard to work 14:52:02 <tellesnobrega> I can do that 14:52:08 <jeremyfreudberg> thanks! 14:52:44 <tellesnobrega> do we have anything else on apiv2? 14:53:08 <jeremyfreudberg> yeah, the fake microversion thing -- but just leave your comments (if any) on the review -- we are running out of metting time 14:53:27 <tellesnobrega> sure, I will take a look this afternoon 14:54:26 <tellesnobrega> #topic Open Discussion 14:54:33 <tosky> two quick notes: 14:54:37 <tellesnobrega> we have about 5 minutes ish 14:54:43 <tosky> - please review zchkun's patch about code sharing in vanilla - I would, but I don't have an easy to use vanilla environment around 14:55:06 <tellesnobrega> tosky, I can do that as well 14:55:20 <tosky> - you may see patches from me in the not-far future about switching the test runner for sahara-dashboard from nose to django runner, as horizon (and most dashboards) did 14:55:25 <tosky> that's it 14:55:48 <tellesnobrega> we will keep an eye on it 14:56:03 <tellesnobrega> zchkun, let me know if you need help with the image for mapr6 14:56:23 <tosky> oh, and if you wonder why https://review.openstack.org/#/c/579114/ didn't merge - it's because it depends on https://review.openstack.org/#/c/571480/ 14:56:48 <jeremyfreudberg> another one: tosky, since you seem to have a cloudera environment -- have you tried s3 there? if not, i will probably try it sometime in August, and call it a Bugfix 14:56:51 <jeremyfreudberg> if needed 14:57:15 <tosky> jeremyfreudberg: I didn't try it yet; I can probably, so I can see how job works 14:57:21 <tosky> a job works 14:58:02 <jeremyfreudberg> yep 14:58:50 <zchkun> tellesnobrega, I am building the image and need to take some more time 14:58:54 <tellesnobrega> jeremyfreudberg, tosky I plan to test s3 as well nad I might need help, watch out for pings on it 14:59:04 <tellesnobrega> zchkun, cool, let me know if you need me 14:59:19 <zchkun> ok, thanks 14:59:34 <tellesnobrega> and we are out of time 14:59:44 <tellesnobrega> great meeting everyone 14:59:52 <tosky> my 3rd meeting in a row :D 15:00:00 <tosky> but yeah, good meeting 15:00:04 <tellesnobrega> you should get some rest 15:00:12 <jeremyfreudberg> yes, lots to do 15:00:12 <tellesnobrega> thanks everyone 15:00:16 <jeremyfreudberg> keep up the good work 15:00:17 <tellesnobrega> see you all next week 15:00:21 <tellesnobrega> yeap 15:00:23 <tellesnobrega> :) 15:00:27 <tellesnobrega> #endmeeting