00:01:30 <ekcs> #startmeeting congressteammeeting
00:01:31 <openstack> Meeting started Thu Dec 15 00:01:30 2016 UTC and is due to finish in 60 minutes.  The chair is ekcs. Information about MeetBot at http://wiki.debian.org/MeetBot.
00:01:32 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
00:01:34 <openstack> The meeting name has been set to 'congressteammeeting'
00:01:45 <ekcs> Hi all. Good to see everyone!
00:02:07 <ekcs> I have the following topics for today:
00:02:08 <ekcs> - Gating
00:02:09 <ekcs> - Ocata-2
00:02:11 <ekcs> - Ocata-3 timeline
00:02:12 <ekcs> - Ocata client release
00:02:12 <ekcs> - Open discussion
00:02:13 <ekcs> Anything else to add?
00:02:27 <masahito> hi
00:02:35 <ramineni_> Hi
00:02:54 <ekcs> hi masahito ramineni_ !
00:03:36 <thinrichs> Hi all
00:03:57 <ekcs> hi thinrichs !
00:04:25 <ekcs> Okay first topic then  = )
00:04:29 <ekcs> #topic gating
00:04:41 <ekcs> Everyone’s fav topic I’m sure.
00:04:51 <ekcs> I think the gate is finally happy after lots work. It seems that when a process/thread interrupts the current test with logging, subunit checksums fail (I still don't understand the exact mechanics).
00:05:10 <thinrichs> Nice work!
00:05:13 <ekcs> Big thanks to ramineni_ for resolving the biggest troublemakers here: https://review.openstack.org/#/c/408439/
00:05:23 <ekcs> Another change here resolved a less frequent source of problem: https://review.openstack.org/#/c/409380/
00:05:33 <ekcs> As far as I can tell, parser checksum issues have not shown up since those two patches.
00:05:37 <ekcs> Any thoughts/comments/questions about gating?
00:06:34 <thinrichs> A quick one: all those node.stop()s
00:06:46 <thinrichs> weren't handled in the destructor for those tests?
00:06:56 <thinrichs> #link https://review.openstack.org/#/c/408439/6/congress/tests/dse2/test_dse2.py
00:08:27 <ekcs> thinrichs: Yes. But afaik destructors may not execute right away. i dunno if python does reference counting or periodic sweep GC.
00:09:13 <thinrichs> Interesting.
00:09:16 <ekcs> on the other hand, I’m not sure we saw the problem if the stops are taken out. ramineni_ do you have any comments?
00:09:36 <ekcs> perhaps we should try taking out the stops just to see.
00:09:41 <ramineni_> Thinrichs: You meant in teardown? Or by default
00:09:52 <thinrichs> teardown
00:10:35 <ramineni_> It's not defined in the file .. And all the nodes created are specific to each test ..so
00:10:49 <ramineni_> Not defined in setup
00:11:34 <ekcs> thinrichs: you mean the python __del__ right? not the test tearDown method.
00:11:35 <thinrichs> I see.  I thought at some point we had tests that registered the nodes they created to be sure teardown could destroy them.  But maybe that was a different test.
00:11:45 <thinrichs> Or __del__ eventually.
00:12:16 <ramineni_> Ya .. not sure it actually stopped the error though
00:12:46 <ramineni_> We had those in API tests
00:13:56 <ekcs> I guess we might do the same thing here so it’s not easy to forget the stop and potentially hit mysterious problems.
00:14:01 <thinrichs> I'll put this in the back of my mind for next time.  Thanks
00:14:37 <thinrichs> I think it's just as easy to forget to register the nodes as it is to stop them.  Unless there were a helper for creating nodes that also registered them.
00:14:48 <ekcs> thinrichs: yup.
00:17:18 <ekcs> ok moving on then.
00:17:35 <ekcs> it’s nice to finally have working gate.
00:17:44 <ekcs> #topic ocata-2
00:17:56 <ekcs> Ocata-2 is due today/tomorrow. I've started doing testing on the latest commit and barring any serious issues will submit a patch to tag this as ocata-2: https://github.com/openstack/congress/commit/9943221dacd6e9215ada891dd3e89fcd6b5fbd21
00:17:56 <ekcs> Any thoughts/comments/questions?
00:20:51 <ekcs> ok moving on then.
00:20:54 <ekcs> #topic Ocata-3 timeline
00:21:15 <ekcs> Looking at the release schedule: #link https://releases.openstack.org/ocata/schedule.html
00:21:16 <ekcs> There is a very compressed timeline between ocata-3 and RC1. So I think we should really frontload our testing to make sure ocata-3 is as close to final release as possible (which is always a good idea anyway). So I think we could aim to get all the Ocata changes in by january 16 and have a full week of testing before Ocata-3 release week.
00:21:16 <ekcs> What do you guys think about that?
00:23:22 <thinrichs> So it's 1 week from Ocata-3 to RC1?
00:23:33 <ekcs> thinrichs: yup.
00:23:44 <ekcs> Joys of a short release cycle!
00:24:11 <thinrichs> Trying to finish up feature work a bit early seems good.
00:25:06 <ramineni_> +1
00:25:12 <masahito> +1
00:25:33 <ekcs> Okay then. I'll remind everyone again as it gets closer.
00:25:41 <ekcs> Moving on if there’s nothing else.
00:25:55 <ekcs> #topic Ocata client release
00:26:10 <ekcs> We have actual changes to the client this time around. So it'll be more important to do thorough testing there too. Does anyone want to spearhead that effort?
00:28:34 <ramineni_> Any major change ?
00:28:56 <masahito> Did we have any change that effects API schema or something?
00:29:22 <masahito> If we have, we need to check the client side.
00:29:54 <ekcs> ramineni_: I think minor changes. But changes nonetheless. removing deprecated oslo.common code. etc. compared to last cycle whee it’s essentially just doc and req changes.
00:30:17 <ekcs> masahito: I don’t think there is any API schema change. anyone know otherwise?
00:31:00 <thinrichs> We had some DB schema change go into the server this time.  Does that impact the client?
00:31:20 <thinrichs> Don't remember which table—handled soft deletes of maybe policies?
00:31:42 <thinrichs> But all that should be hidden by the API
00:31:56 <thinrichs> Have any of the APIs changed?
00:32:15 <thinrichs> I don't remember any
00:32:46 <ekcs> thinrichs: oh hmm good point. How does the client access the DB anyway?
00:33:01 <thinrichs> thru the API (and only thru the API)
00:33:10 <ekcs> got it.
00:33:20 <ekcs> no API changes AFAIK.
00:34:55 <ekcs> Anyway I don’t mean we need any extraordinary testing. Just saying we need to run through all the commands and check that nothing broke. Anyway we can circle back on that at a later day.
00:35:07 <ekcs> any other thoughts/comments/questions?
00:38:00 <ekcs> ok then moving on.
00:38:01 <ekcs> #topic open discussion
00:38:01 <ekcs> Feel free to bring up anything you want to discuss or give a status update.
00:38:22 <thinrichs> Attended the policy meeting this morning again.
00:38:48 <thinrichs> One of the interesting bits that we haven't thought much about is cross-project (cross-tenant) concerns.
00:39:12 <thinrichs> Seems that some companies have 1000s of projects/tenants.
00:39:24 <thinrichs> Each new user has 1+ tenant/project of which they are the admin.
00:39:54 <thinrichs> Each resource is owned by exactly one tenant/project (though it seems there are some exceptions to that rule, like networks).
00:41:55 <thinrichs> When an API call comes in oslo-policy knows about this rule and checks if the object impacted by the API call belongs to the same project that the user belongs to.
00:42:45 <thinrichs> What I don't understand right now is whether project1 can grant rights for the object it owns to users of project2.
00:42:59 <thinrichs> But that was one of the use cases that came up during the meeting.
00:43:35 <thinrichs> For us, the owner of the resource is just a property of that object, and so we'd have no trouble granting rights
00:43:53 <thinrichs> for an object in project1 to users in project2.
00:44:17 <thinrichs> But doing that requires having *global* visibility into the objects/users of all the different projects.
00:44:52 <thinrichs> One thing that Ruan did with the Moon policy-engine (don't have link handy) is enable each project to have its own policy.
00:45:51 <aimeeu> #link https://wiki.opnfv.org/display/moon/Moon
00:45:52 <thinrichs> One thing we might want to consider is understanding how Congress works in a multi-tenant world.
00:45:57 <thinrichs> aimeeu: thanks!
00:47:14 <thinrichs> We support multiple policies.  Can we support each project/tenant writing their own policies?  Would we still want global-policies? How would the project/tenant policies and the global policies interact?
00:47:36 <ekcs> hmm very interesting questions to consider.
00:48:12 <thinrichs> Sorry that was so long.
00:48:46 <ekcs> thanks for the update thinrichs !
00:49:25 <ramineni_> Just one more item would like to discuss .. Any thoughts of making congress dashboard separate project??
00:49:31 <ekcs> sounds like a great topic to think about and be prepared to drill into at the PTG.
00:49:56 <aimeeu> I think making the dashboard a separate project is a great idea
00:50:17 <thinrichs> Makes sense to me
00:52:16 <ramineni_> We could consider for ocata ?
00:52:52 <ramineni_> Or its too late?
00:52:57 <aimeeu> Is it a lot of work to set up a new project?
00:53:01 <ekcs> are there good examples we can follow? like other projects having a dashboard project?
00:53:26 <ramineni_> Many projects do
00:53:34 <thinrichs> Not sure how hard it is to setup a new project; haven't done that myself.
00:54:06 <ramineni_> I have mentioned in mailing list .. Let me get the link
00:54:40 <ekcs> could that be done in a way that doesn’t affect congress project? like start the process of setting up the project, but without removing from congress just yet.
00:54:52 <ramineni_> http://docs.openstack.org/developer/horizon/plugin_registry.html
00:55:15 <aimeeu> #link http://docs.openstack.org/infra/manual/creators.html
00:55:32 <ekcs> that way we don’t run the risk of the new thing not being ready and compromising ocata funcitonaliy of congress.
00:59:25 <ekcs> a minutes left. would someone like to figure out a plausible plan and timeline for creating doshboard project?
00:59:34 <masahito> Just question. Policy dashboard for a tenant seems to be ok, butCan current congress handle a tenant/project policy rule now?
01:00:32 <ekcs> time’s up for the meeting. why don’t we continue on #congress for those interested.
01:00:33 <aimeeu> ekcs: I can look into what it will take to create a separate dashboard project since I, um, asked for it. ;)
01:01:06 <ekcs> thanks aimeeu !
01:01:07 <masahito> ok, bye
01:01:25 <ramineni_> Thanks aimeeu
01:01:32 <ekcs> ok see you all next time or on #congress !
01:01:36 <ekcs> #endmeeting