17:12:24 <davidlenwell> #startmeeting refstack
17:12:24 <openstack> Meeting started Thu Apr 17 17:12:24 2014 UTC and is due to finish in 60 minutes.  The chair is davidlenwell. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:12:25 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:12:27 <openstack> The meeting name has been set to 'refstack'
17:12:32 <davidlenwell> roll call?
17:12:39 <catherineD> Here
17:12:45 <tedchang> here
17:12:57 <fcarpenter> morning
17:13:05 <davidlenwell> morning everyone
17:13:08 <praveen_dell> good evening
17:13:24 <davidlenwell> #topic summit session
17:13:57 <davidlenwell> catherineD: will you be able to attend the openstack summit?
17:14:16 <catherineD> Last I check the session is pre-accepted?
17:14:30 <davidlenwell> yes .. I think it will go through
17:14:35 <catherineD> we can only travel if it is accepted ..
17:14:46 <davidlenwell> okay .. so we'll know soon enough
17:15:39 <catherineD> yes .. only then we can make travel request
17:16:03 <davidlenwell> costs more that way.. I personally live in hope that it will be ..
17:16:07 <davidlenwell> but I was going to the summit anyways
17:16:31 <catherineD> good good, I think Rocky is too ..
17:16:38 <davidlenwell> okay moving right along
17:16:52 <praveen_dell> i would like to attend the summit
17:17:34 <davidlenwell> praveen_dell: tell rob you must attend ;)
17:17:45 <davidlenwell> #topic execute_test madness
17:18:28 <davidlenwell> So I was hoping for either more pushback on what I did in execute_test or a working tcup build from rob.. I guess he's going to work on it tomorrow
17:19:13 <catherineD> Last status is Ted waits waiting for davidlenwell: before making more change to execute_test?
17:19:40 <davidlenwell> catherineD: I posted a review of it and asked you gusy to review it durring the last meeitng
17:19:51 <catherineD> you understand what I mean :-)
17:20:23 <davidlenwell> since it was blocking rob from his work I merged the code
17:20:45 <davidlenwell> tedchang: did you get a chance to review it .. it was your code I started off with
17:21:07 <catherineD> They did
17:21:15 <davidlenwell> he didn't
17:21:27 <tedchang> both me and ray left comments on it
17:22:18 <davidlenwell> only on docker_buildfile.template  not in the execute_test code it self
17:22:29 <davidlenwell> that actually is a good point ..
17:22:48 <davidlenwell> the docker_buildfile.template pulls the script from refstack .. that is one thing i want to change ..
17:22:53 <catherineD> ok so the code is merged ..
17:23:08 <tedchang> execute_test I looked but I have to try to really make sure it will work
17:23:14 <davidlenwell> I'd like for it to get the script from git rather than from an http call .. for one .. an http call is very insecure
17:23:40 <davidlenwell> to wget a file from a url then execute it leaves us open for man in the middle attacks
17:24:28 <davidlenwell> I don't forsee execute_test changing much in the future once its working .. so I don't want to keep this model ..
17:25:12 <davidlenwell> if you look at what I did I started to make execute_test an installable package with tempest as a dependancy
17:25:53 <davidlenwell> what I would love to do is get it into its own repo called "refstack-tester" or something and get it into pypy so that it can just be pip installed on testing nodes
17:26:38 <davidlenwell> what do you guys think about that ?
17:27:53 <catherineD> Here is Raymond last comment "The main purpose of using the get-script api is to make sure the script and the dockerfile are of the same version. Getting the script from github cannot guarantee they are of the same release."
17:28:38 <davidlenwell> that is a valid argument
17:28:50 <davidlenwell> but the api needs to be version locked moving forward anyways
17:29:18 <davidlenwell> so that the live copy of refstack.org gets api/v1 api/v2 and so forth ..
17:29:40 <davidlenwell> because refstack.org's api needs to be able to accept data from any version of the tester
17:30:25 <davidlenwell> I understand thats hard to do when things are changing rapidly .. but things can't keep shifting as much as they will int he beginings of a project
17:31:12 <catherineD> I have no problem with that ..... We need to focus ... at this point the most important thing is to collect data .. agree?
17:31:20 <davidlenwell> agreed
17:32:18 <catherineD> so what we have so far .. we are able to collect some data ... how do we enable others to collect data?  are they willing to just run what we have that is working ?
17:32:24 <s3wong> Is this the Neutron group-policy meeting?
17:32:34 <banix> s3wong: no
17:32:36 <davidlenwell> s3wong: no this is refstack
17:32:37 <banix> alt
17:33:16 <banix> s3wong: we are at the meeting-alt channel
17:33:23 <s3wong> banix: Thanks
17:33:32 <davidlenwell> catherineD: so what we have so far doesn't allow that currently .. if rob can finish what he is working on we'll be able to collect data from anyone
17:34:12 <catherineD> We have been collecting data via Refstack path ..
17:34:34 <davidlenwell> catherineD: but that data is not going back to refstack.org's central database
17:34:52 <catherineD> the that is what we need to work on ...
17:34:54 <davidlenwell> we're missing the publish results part of the story
17:35:11 <catherineD> make sure that data can be sent back to centrol Refstack
17:35:35 <catherineD> should we concentrated on the send back center Refstack ?
17:35:45 <davidlenwell> possibly ..
17:36:03 <catherineD> Raymond had a TCUP version that is working ...
17:36:17 <catherineD> it may not be the final version .. we all understand that ...
17:36:32 <davidlenwell> are you reffering to the shell scripts ?
17:37:21 <catherineD> I refer to this one https://review.openstack.org/#/c/84320/
17:37:45 <davidlenwell> yeah .. that one
17:37:51 <catherineD> this is a very quick thing that we put together just for data collection ... understand that it may no be final ..
17:38:05 <praveen_dell> i have tested existing tcup,it only builds the docker container and exports openstack properties on to docker container
17:38:16 <catherineD> I really want to see some other data ...
17:38:56 <davidlenwell> catherineD: so do I ..
17:39:16 <catherineD> we can focus on the send back data ... I have  data that I collect from Havana (all in one) env that I can pulbish
17:39:30 <davidlenwell> so rob tells me he'll have time tomorrow to finish what he was working on .. the tcup.py aproach
17:40:24 <davidlenwell> that is .. sourcing an openstack.rc file that a user gets from the cloud they want tested
17:40:34 <catherineD> np .. just that we need data ... I found that there are more configuration in tempest.config that we need to consider ... that is why I would like to see more data from other env
17:40:45 <davidlenwell> then kicking off a script that grabs the already built container and runs it with the env variables
17:40:59 <catherineD> source openstack.rc is definitetly not enough ..
17:41:12 <davidlenwell> it should be able to auto discover the rest
17:41:16 <catherineD> one example ... network ..
17:41:56 <davidlenwell> upstream tempest is close to having the networking auto discovery built ..
17:42:05 <catherineD> if you env has more than one network and your network name is not named "private" or "public"  so need to confgure those parameters ...
17:42:26 <davidlenwell> not sure how long it will take for that code to make it into havana stable
17:42:38 <catherineD> discovery is also a problem .. let say I discover 10 networks then which one should I use?
17:42:51 <davidlenwell> that is going to require some thought ..
17:43:06 <davidlenwell> lets get back to a network / config discussion offline
17:43:12 <davidlenwell> we're low on time
17:43:19 <catherineD> yes .. a lot more configuration needed as I start to collect data
17:43:27 <davidlenwell> okay . we're running low on time so I am going to move to another topic
17:43:46 <catherineD> sure
17:43:47 <davidlenwell> #topic #infra testing
17:44:19 <davidlenwell> So while at pycon over the weekend I had a chance to chat face to face with jim blair
17:44:32 <davidlenwell> he's ptl of infra .. the group we want to host this thing long term
17:44:58 <davidlenwell> and he sorta shattered some of my dreams about how I wanted to trigger tests using pure openstack things..
17:45:06 <catherineD> praveen_dell: may be take a look at this one for some idea https://review.openstack.org/#/c/84320/ of how we run test ... Let's move to the next subject
17:45:19 <davidlenwell> but then explained a workable solution that might be a better idea anyways
17:45:36 <praveen_dell> @catherineD sure
17:45:59 <davidlenwell> that is essentially to just use the infra system to run tests
17:46:50 <catherineD> davidlenwell: that is refstack will be hosted on infra system?
17:46:51 <davidlenwell> without explaining in detail how infra works ..
17:47:02 <davidlenwell> catherineD: yes .. it will
17:47:12 <davidlenwell> but even before it is .. they can host the testing env
17:47:24 <davidlenwell> they have a thing called nodepool
17:47:28 <davidlenwell> and another thing called zuul
17:47:57 <davidlenwell> nodepool just makes a bunch of ready to run .. one time use vm's for running tests .. their config is passed to them with a yaml file
17:48:27 <davidlenwell> when they boot up they ask zuul for jobs.. zuul gives them a job and then keeps track of its progress
17:48:39 <davidlenwell> So we can piggy back on that
17:48:52 <davidlenwell> create our own node type for nodepool
17:49:22 <davidlenwell> and im paraphrasing here .. but essentially it will keep an army of ready to run single use vm's at our disposal
17:49:40 <davidlenwell> that we can then trigger and let run and report home with results
17:52:37 <davidlenwell> anyways .. long story short .. my original gearman spec is close to what we'll end up with but not exactly .. and I'll amend the spec to more closely match that
17:52:53 <catherineD> cool
17:53:30 <davidlenwell> #topic open discussion
17:54:10 <davidlenwell> is there anything anyone needs to talk about ?
17:54:52 <catherineD> so we will work on the send data back to code  so we can collect data ....  agree?
17:54:54 <davidlenwell> catherineD: I feel like the networking discussion isn't going to be an easy solution .. perhaps we'll need to make this user entered feilds to start with
17:55:08 <catherineD> yes exactly ...
17:55:23 <catherineD> that is why it is important to have more people test to discover more ...
17:55:43 <davidlenwell> catherineD: since refstack.org already has a public method for accepting data .. we just need to make a call out to api thing on the local refstack that can send things to it
17:55:56 <catherineD> i have the same thinking as yours that we may need more user data enter than we would like to ...
17:56:19 <catherineD> ok will do that to call the pulbikc methoid ....
17:56:41 <davidlenwell> catherineD: can you write a spec before writing any code please ?
17:56:42 <catherineD> At this time data is what we need ...
17:57:07 <davidlenwell> the public method might have to be change .. the database might need some alteration
17:57:13 <davidlenwell> and I just want it to be thought through
17:57:32 <catherineD> we will but we do not see writing any more code before discussion with the team ...
17:57:36 <davidlenwell> the questions in the spec template are designed to take out the guessing on the impact of changes
17:58:07 <catherineD> ic
17:58:11 <catherineD> will do
17:59:21 <davidlenwell> okay .. I look forward to seeing that .. I'll be back at piston hq full time next week .. So my hours will be more regular and we'll see more commits out of me
18:00:07 <catherineD> great! maybe we need an other f-2-r
18:00:12 <catherineD> f2f
18:00:20 <davidlenwell> yeah .. im open to that
18:00:41 <davidlenwell> we should plan for after the 1st of the month tho .. piston is moving offices right now
18:00:51 <davidlenwell> #endmeeting