19:01:20 <davidlenwell> #startmeeting refstack
19:01:21 <openstack> Meeting started Mon Dec 22 19:01:20 2014 UTC and is due to finish in 60 minutes.  The chair is davidlenwell. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:01:22 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
19:01:24 <openstack> The meeting name has been set to 'refstack'
19:04:10 <hogepodge> o/
19:04:18 <hogepodge> sorry I'm late.
19:04:29 <davidlenwell> it happens
19:04:29 <zehicle> o/
19:06:54 <sslypushenko_> o/
19:07:32 <zehicle> agenda?
19:08:47 <davidlenwell> let first get into the mid-cycle meetup
19:09:01 <davidlenwell> #topic mid-cycle meetup
19:11:27 <davidlenwell> rocky doesn't seem to be here .. but in our email threads it would apear that she has successfully secured us some space and lunch for the dates of  jan 12,13,and 14
19:12:55 <davidlenwell> we'll get her to send a group email informing us of the details so folks can get hotels and travel arrainged
19:13:12 <davidlenwell> sslypushenko_: we'll also make sure we setup a google hangout
19:13:28 <davidlenwell> does anyone have anything to ad?
19:13:34 <sslypushenko_> Sounds good!
19:14:05 <davidlenwell> okay .. next topic I want to cover is the api
19:14:17 <davidlenwell> sslypushenko_: have you started on anything or are you blocked?
19:14:30 <vladiskuz_> i started working on api's prototype
19:14:39 <vladiskuz_> on pecan
19:14:47 <davidlenwell> excelent!
19:14:51 <sslypushenko_> But we need to discuss db model
19:14:59 <davidlenwell> sure
19:15:06 <davidlenwell> lets discuss it
19:15:08 <sslypushenko_> I put some comments on your commit
19:15:12 <davidlenwell> okay
19:15:15 <davidlenwell> I'll look at that
19:15:52 <sslypushenko_> I think we should add timestamp field and metadata field to database
19:16:19 <davidlenwell> can you explain what you mean more by a metadata field?
19:17:00 <sslypushenko_> I guess it should be json blob stored as string
19:17:14 <davidlenwell> what data would it store?
19:18:11 <sslypushenko_> If we need to some data in future we can store it in metadata without db migration.
19:18:32 <zehicle> since the test meta data is open ended, it makes sense to be able to store whatever we get from the tests
19:18:41 <sslypushenko_> It will helpfull while we developing our service
19:18:42 <davidlenwell> I think thats a slippery slope .. and that kind of data isn't efficiently searchable
19:18:52 <sslypushenko_> sure
19:19:00 <sslypushenko_> it is only for storing
19:19:06 <davidlenwell> the other table is strucuted to be metadata
19:19:13 <davidlenwell> we could just put extra stuff in it
19:19:48 <zehicle> that way we would not lose any of the meta data that
19:19:48 <sslypushenko_> It we need to search we will migrate db
19:19:50 <zehicle> s being added to the tests
19:20:15 <davidlenwell> yeah.. I think zehicle and I are on the same page right now
19:20:48 <zehicle> ??
19:21:12 <sslypushenko_> I see only test results field
19:21:42 <sslypushenko_> We can add metadata to it. But it is not perfect solution
19:22:24 <sslypushenko_> Also 512 - it is not enough size for test result
19:22:29 <zehicle> don't we track results per test?
19:22:41 <davidlenwell> I think you've missunderstood the table structure
19:22:49 * zehicle has not looked at the model for long time
19:23:05 <hogepodge> So Test is one run, and TestResults is the result of a single test?
19:23:12 <davidlenwell> no no
19:23:38 <davidlenwell> refstack only cares about pass/fail .. we don't store the full output of a test
19:23:40 <hogepodge> One test has many testresults?
19:24:12 <davidlenwell> so the key / value pair structure is ideal .. we can store the test uuid / pass-fail in those columns
19:24:29 <davidlenwell> so one run might have 1000 records in that table
19:24:43 <davidlenwell> it leaves the number of tests flexible
19:24:51 <davidlenwell> and insures we don't have to worry about feild size
19:25:12 <sslypushenko_> It this case I agreed
19:25:25 <davidlenwell> well thats what the intention of that table design is
19:25:38 <davidlenwell> its not one record per run ..
19:25:51 <davidlenwell> its one record per test .. one run will have many many records ..
19:26:17 <sslypushenko_> Now I'm understand
19:26:31 <davidlenwell> so do you still have the same objections to my table design?
19:26:42 <zehicle> so, we're talking about meta data in the test record fields?  if so, that aligns w/ having flexible meta data in the test
19:26:59 <davidlenwell> no
19:27:13 <davidlenwell> why would we store the tests meta data?
19:27:21 <sslypushenko_> we can put matadata and timestamp in key/value table... I think it will work
19:27:35 <davidlenwell> what metadata do you want to store?
19:27:39 <davidlenwell> seems excesive
19:28:12 <davidlenwell> I'm fine with adding a time stamp
19:28:16 <davidlenwell> seems logical
19:28:27 <davidlenwell> but why do we need meta on every test result record?
19:28:44 <zehicle> because we don't have control over what's added
19:28:55 <sslypushenko_> every result - no
19:28:58 <zehicle> and it gives us a way to keep the data that's there as we figure out what to do
19:29:14 <sslypushenko_> but we should can if we need one
19:29:21 <zehicle> it's possible that devs will add meta data to tests that we want
19:29:32 <zehicle> and that it will need time to settle
19:29:48 <sslypushenko_> Now I'm ok with db schema
19:29:52 <davidlenwell> I think its out of our scope to store all the meta for all the tests we run attached to every result
19:29:52 <zehicle> so, keeping it open allows us to capture data even it's final
19:30:01 <zehicle> but if everyone else is OK, then I'm fine
19:30:16 <hogepodge> Would it be better to add metadata as a separate table entry that points back to the test? That way it's not implied to be a key part of the parent table.
19:30:18 <zehicle> is there any metadata there now?
19:30:19 <davidlenwell> I think sslypushenko_ just missunderstood the original intent of the design
19:30:29 <sslypushenko_> Yep
19:30:47 <davidlenwell> we don't need to store it at all hogepodge its in tempest in the code .. doesn't make sense for us to maintain meta in our database ..
19:30:55 <davidlenwell> but i think this is another issue
19:31:06 <sslypushenko_> I think only secondary table name should be corrected, I'm right?
19:31:31 <davidlenwell> sslypushenko_:  are you suggesting we add the word "meta"
19:31:32 <davidlenwell> ?
19:32:09 <sslypushenko_> I suggest any data what can be connected with test run
19:32:20 <sslypushenko_> User comment for example
19:32:27 <davidlenwell> sure
19:33:21 <davidlenwell> lets not bike shed any further on this topic ..
19:33:38 <davidlenwell> sslypushenko_: please reach our in channel if you want to discuss the db model further
19:34:04 <sslypushenko_> okay
19:34:37 <sslypushenko_> You can land the schema and we can start with api
19:34:43 <davidlenwell> okay
19:34:51 <davidlenwell> can you go change your -1
19:36:01 <sslypushenko_> Table name is ok&
19:36:09 <sslypushenko_> Table name is ok?
19:36:09 <davidlenwell> thanks
19:36:17 <davidlenwell> I don't have a problem with it
19:36:23 <sslypushenko_> in 33 line
19:36:37 <davidlenwell> one sec
19:36:58 <davidlenwell> oh.. pep8 requires class names to be upper camel case
19:37:01 <sslypushenko_> I think it sure be differ from 'test'
19:37:18 <sslypushenko_> Or I'm wrong?
19:37:39 <davidlenwell> oh.. Ididn't catch that
19:37:41 <davidlenwell> I'll fix it
19:37:58 <davidlenwell> I'll push a new revision today
19:38:01 <sslypushenko_> I will wait for your update
19:38:26 <sslypushenko_> I guess that is it with schema
19:38:38 <davidlenwell> great
19:38:47 <sslypushenko_> Can we discuss test uuid spec?
19:39:28 <davidlenwell> its in process.. I spent a lot of last week arguing with folks about some details.. I will also push it live no later than lunch tomorrow .. I still have an outstanding issue
19:39:52 <davidlenwell> I know people are waiting on this.. but because this spec is actually for another project all together we can't just bulldoze in with our requirements ..
19:39:57 <davidlenwell> it effects a much larger picture
19:40:04 <sslypushenko_> I have almost done script which add decorator "test.uuid(uuid='blahblahblah')" to all tempest tests
19:40:08 <hogepodge> this needs to be communicated.
19:40:20 <hogepodge> It came up in the defcore meeting. "Where are uuids on tempest tests."
19:40:51 <davidlenwell> while I understand that.. defcore doesn't get to tell q/a how to write there code or how to structure meta data in there tests
19:40:54 <sslypushenko_> In comparison with docstring-way this code looks much solid
19:40:56 <davidlenwell> there is a process here
19:40:59 <hogepodge> It is important to many people across many teams, and we've been silent on it.
19:41:14 <hogepodge> What's the process? I'm trying to understand.
19:41:30 <davidlenwell> hogepodge: lets discuss that on our upcomming call
19:41:47 <davidlenwell> if there is nothing else to discuss I am going to push this issue..
19:42:08 <zehicle> can someone post a link to the the latest spec on it?
19:42:35 <davidlenwell> it will be in gerrit shortly zehicle I'll make sure you guys all know about it when its up
19:43:14 <davidlenwell> okay then.. if nobody has anything different to dicsuss I am going to end the meeting
19:43:41 <zehicle> I'd like to talk about cleaning up gerrit reviews
19:43:44 <zehicle> that are out of date
19:43:50 <davidlenwell> okay
19:44:07 <zehicle> "Updated Swift section of havana sections." was superceded
19:44:08 <davidlenwell> I went through and abandonded the wip reviews I had going fo the api
19:44:26 <zehicle> "Adds a description for each capability in capabilities.json" also was superceded
19:44:41 <zehicle> the details of each made it into a different patch
19:44:44 <davidlenwell> can you have the original devs click the abandon button ?
19:45:07 <davidlenwell> otherwise they'll just self abaondon eventually
19:45:12 <davidlenwell> I don't remember what the time out is
19:45:28 <davidlenwell> as you know I was pinging you about those last week
19:46:10 <zehicle> yy
19:46:36 <notmyname> davidlenwell: (pedant) patches no longer auto-abandon. you gotta have the original devs to it. or "core" can do it
19:46:57 <zehicle> notmyname, one is yours - can you do it
19:47:01 <davidlenwell> if we cannot get the original devs to do it I'll abandon them
19:47:03 <notmyname> zehicle: link?
19:47:14 <zehicle> #link https://review.openstack.org/#/c/122573/
19:47:54 <notmyname> done
19:48:23 <zehicle> harder to get Troy back in
19:48:34 <zehicle> troytoman, is on!
19:49:02 <zehicle> troytoman, can you abandon https://review.openstack.org/#/c/110399/
19:50:11 <davidlenwell> okay .. anything else we need to cover?
19:50:58 <sslypushenko_> Nothing from my side
19:51:07 <zehicle> did we get the pull for the static page in?
19:51:29 <zehicle> that would allow hogepodge to work the infra side for CI
19:51:32 <davidlenwell> no.. I'll add that to my list this week
19:51:44 <davidlenwell> we still can't do that until we have the api
19:51:49 <davidlenwell> thats in process
19:51:51 <davidlenwell> if anyone needs anything hit me on #refstack
19:51:54 <davidlenwell> #endmeeting