16:04:57 <kgriffs> #startmeeting marconi 16:04:58 <openstack> Meeting started Mon Aug 12 16:04:57 2013 UTC and is due to finish in 60 minutes. The chair is kgriffs. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:04:59 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:05:01 <openstack> The meeting name has been set to 'marconi' 16:05:04 <flaper87> \o/ 16:06:28 <flaper87> toc toc ? 16:06:29 <kgriffs> let me pull up the agenda 16:06:30 <flaper87> :D 16:06:37 <kgriffs> any actions from last time? 16:06:53 <flaper87> Yeah, there were a couple of actions w.r.t tests 16:06:54 <kgriffs> I didn't see a summary posted on the wiki... 16:06:57 <flaper87> IIRC 16:07:06 <malini> we had a heavy test discussion & I added it to the bp 16:07:23 <malini> https://blueprints.launchpad.net/marconi/+spec/refactor-system-tests 16:07:36 <cppcabrera> I updated the summary a few days ago. I also noticed it was missing, kgriffs. 16:07:46 <cppcabrera> #link http://eavesdrop.openstack.org/meetings/marconi/2013/marconi.2013-08-02-16.02.html 16:08:03 * kgriffs gives cppcabrera a pop-tart 16:08:10 <cppcabrera> :) 16:08:13 <flaper87> arrgh, that's my fault 16:08:16 <flaper87> sorry guys! 16:08:23 <flaper87> I completely forgot ot update it 16:08:29 <malini> one of the ACTION is ' kgriffs to buy pop-tarts' ;) 16:08:33 <kgriffs> heh 16:08:36 <cppcabrera> No worries. You've got backup, flaper87. ;) 16:08:43 <kgriffs> #topic review actions from last time 16:09:05 <kgriffs> ok, so any items we need to carry forward? 16:09:16 <cppcabrera> Both. 16:09:22 <flaper87> yeah, 1 and 2 16:09:27 <malini> Action # 3 & 4 too 16:09:47 <cppcabrera> Client dev got very little love this go around. 16:09:51 <kgriffs> you mean, everything under topic 3? 16:09:57 <kgriffs> (c,e,f,g) 16:10:06 <kgriffs> oic 16:10:23 <malini> f is complete 16:10:40 <malini> (its in the bp + etherpad) 16:11:01 <kgriffs> 2.b., 2.d, 3.c, 3.e 16:11:07 <kgriffs> all those need to be carried forward? 16:11:31 <flaper87> (actions items are enumerated bellow) 16:11:32 <flaper87> kgriffs: yup 16:11:42 <kgriffs> ok 16:12:01 <kgriffs> #action cppcabrera and flaper87 to turn apiclient-marconi etherpad into bps/bugs 16:12:05 <malini> 2.d is abadoned 16:12:22 <kgriffs> #action malini to update the etherpad and start working on the re-factor 16:12:31 <kgriffs> #action malini to make functional tests run with tox 16:12:37 <kgriffs> does that look right? 16:12:45 <malini> yes 16:12:47 <flaper87> it does to me 16:12:49 <cppcabrera> yup 16:12:58 <kgriffs> ok, moving right along 16:13:02 <cppcabrera> +1 16:13:02 <kgriffs> #link https://wiki.openstack.org/wiki/Meetings/Marconi#Agenda 16:13:14 <kgriffs> #topic poposal to deprecate Trello 16:13:31 * flaper87 is ready to vote 16:13:42 <kgriffs> Any question/comments before we vote? 16:13:45 <zyuan_> which one we use? 16:14:06 <cppcabrera> Launchpad (LP) 16:14:10 <kgriffs> today we use both Trello and Launchpad, but it had become a real waste of time (IMHO) to keep them in sync 16:14:17 <Sriram> +1 16:14:25 <zyuan_> launchpad has card? 16:14:39 <flaper87> zyuan_: nope, just issues and blueprints 16:14:41 <kgriffs> Trello seems really helpful when planning the next release, but after that becomes hard to maintain since it has to sync with Launchpad. 16:15:14 <kgriffs> also, we are the only project i know of using Trello at the moment. I would rather focus efforts on improving Launchpad and/or helping infra migrate to something like phabricator. 16:15:27 <zyuan_> it's somewhat not obvious to see who is working on which bp/bug 16:15:51 <flaper87> zyuan_: if we use assignee fields correctly, it becomes obvious 16:16:00 <flaper87> we haven't done that because we were focusing on trello 16:16:15 <kgriffs> zyuan_: agreed, a lot of people are not satisfied with Launchpad, but aint nobody got time to keep Trello and LP in sync, and sometimes comments and stuff end up in one place and not the other. 16:16:15 <malini> I like the moving thru column part of trello 16:16:26 <zyuan_> when i say not obvious i mean... the bugs page does not give you such info 16:16:30 <zyuan_> you have to open each 16:16:31 <malini> helps me track all work has tests associated 16:16:40 <kgriffs> how about we try it for a few weeks and see? 16:16:46 <kgriffs> (try not using Trello) 16:16:57 <zyuan_> cards are different. you don't need to open it to know lots of things 16:17:38 <flaper87> We'll have to move to launchpad anyway, that's what OS uses and that's what TC and everyone will use to get status of the project 16:17:43 <kgriffs> an alternative would be to take a day or two and write a "facelift" interface on top of launchpad that makes it more grokkable at a glance 16:18:00 <flaper87> Moving forward with trello wont replace launchpad, anyway 16:18:01 <zyuan_> ha ha ha ha 16:18:13 <flaper87> kgriffs: lol 16:18:20 <Sriram> :P 16:18:33 <flaper87> zyuan_: I agree with you :/ 16:18:38 <kgriffs> so, if anyone is looking for a totally sweet project for their next hackathon... 16:18:51 <amitgandhi> im surprised no one has done it yet 16:19:11 <zyuan_> lunchpad - trello integration? 16:19:19 <zyuan_> or cards for launchpad? 16:19:21 <flaper87> to do that, you'd have to mess with launchpad's OpenId and permissions, THAT is not fun :D 16:19:23 <kgriffs> no, just making launchpad prettier 16:19:41 <flaper87> ok, vote ? 16:19:53 <zyuan_> like bugs in many views? 16:20:16 <kgriffs> yeah, better views. And maybe a nicer font or something. :p 16:20:24 <Sriram> ok lets vote. 16:20:46 * cppcabrera ready to vote 16:21:35 <kgriffs> #startvote Should we try freezing Trello and start using Launchpad exclusively? Yes, No 16:21:36 <openstack> Begin voting on: Should we try freezing Trello and start using Launchpad exclusively? Valid vote options are Yes, No. 16:21:37 <openstack> Vote using '#vote OPTION'. Only your last vote counts. 16:21:52 <cppcabrera> #vote Yes 16:21:56 <flaper87> #vote Yes 16:22:00 <amitgandhi> #vote Yes 16:22:01 <Sriram> #vote Yes 16:22:05 <kgriffs> #vote Yes 16:22:09 <megan_w> #vote yes 16:22:32 <zyuan_> no solution in mind; abstain 16:22:39 <kgriffs> noted 16:22:58 <kgriffs> I'll give it 10 seconds in case anyone wants to object 16:23:23 <kgriffs> #endvote 16:23:24 <openstack> Voted on "Should we try freezing Trello and start using Launchpad exclusively?" Results are 16:23:25 <openstack> Yes (6): megan_w, kgriffs, cppcabrera, amitgandhi, Sriram, flaper87 16:24:23 <zyuan_> one thing i like about launchpad is Link to CVE 16:24:24 <kgriffs> #action kgriffs,flaper87, et al. to stop using Trello and focus efforts on getting/keeping Launchpad up to date 16:24:31 <zyuan_> but we never used that yet :) 16:25:10 <zyuan_> hope we don't use that in the future LOOL 16:25:18 <kgriffs> #action cppcabrera to assist in getting cards on Trello ported to Launchpad for those that aren't in sync. 16:25:37 <kgriffs> #topic triage bugs 16:25:51 <kgriffs> #link http://goo.gl/RjZJCw 16:26:07 <cppcabrera> +1 16:26:19 <kgriffs> malini has identified these as the only two critical bugs at the moment. Actually, I think there is one more pending... 16:26:26 <cppcabrera> There is. It's... 16:26:29 * cppcabrera searches for link 16:26:31 <kgriffs> cppcabrera: retry on duplicate key error? 16:26:36 <cppcabrera> #link https://bugs.launchpad.net/marconi/+bug/1211386 16:26:38 <cppcabrera> Yes 16:26:46 <cppcabrera> The regex match fails. 16:26:53 <cppcabrera> Yielding a 503 16:27:09 <kgriffs> has this been independently verified? 16:27:47 <malini> alejandro verified it, rt? 16:27:49 <cppcabrera> It's been noted by oz_akan and malini while load testing. I'm attempting to repro locally. 16:28:04 <flaper87> mhh, I wonder wether the buffer or the s/queue_id/queue_name/ patches could have caused that 16:28:05 <amitgandhi> cppcabrera reported it 16:28:19 <zyuan_> the releasing claim one, i can't repo it with unit test. i may got something wrong... 16:28:41 <kgriffs> you have to do parallel requests to the same queue 16:28:43 <malini> amitgandhi: cppcabrera reported what oz_akan & me saw in load tests. 16:28:48 <kgriffs> the error is with duplicate markers 16:29:04 <kgriffs> I'd like to mark this one as critical. Any objections? 16:29:07 <Sriram> i'll try to repro it as well. 16:29:14 <malini> amitgandhi: So its not really independent verification 16:29:15 <flaper87> kgriffs: go for it 16:29:56 <kgriffs> ok folks, as soon as we get a reliable repro on this, let's get it fixed 16:30:04 <kgriffs> (pls update the "status") 16:30:09 <flaper87> malini: what version of mongo are you guys using? 16:30:40 <kgriffs> #action malini to verify/repro https://bugs.launchpad.net/marconi/+bug/1211386 16:30:40 <malini> oz_akan_ 16:30:40 <cppcabrera> I'm using v2.4.1 (latest is 2.4.5, I believe) 16:31:18 <malini> kgriffs: this bug we found in load test & we have a reliable repro with tsung 16:31:25 <kgriffs> malini: mostly I just want to make sure we can reproduce this on demand; otherwise how can we prove we "fixed" it? 16:31:30 <kgriffs> sweet 16:31:39 <kgriffs> can you update the status accordingly? 16:31:40 <malini> it shows up on all tsung runs 16:31:52 <oz_akan_> 2.4.5 16:32:18 <kgriffs> moving on 16:32:20 <malini> I changed it to confirmed 16:32:23 <kgriffs> is this one confirmed yet? 16:32:24 <kgriffs> https://bugs.launchpad.net/marconi/+bug/1210662 16:32:28 <kgriffs> #link https://bugs.launchpad.net/marconi/+bug/1210662 16:33:14 <amitgandhi> i reported that. either the spec needs to be corrected, or the code needs to be fixed 16:33:50 <amitgandhi> zyuan is contesting that its a bug and prefers the current approach (which conflicts with the spec) 16:33:52 <kgriffs> I think it's a bug 16:34:00 <zyuan_> i don't think it's bug... 16:34:02 <kgriffs> wait 16:34:09 <kgriffs> how does it conflict with the spec? 16:34:13 <zyuan_> it's spec has a glitch 16:34:25 <zyuan_> spec says age need to be reset 16:34:42 <zyuan_> while current age always reflect the "real" age -- time lived 16:34:45 <kgriffs> isn't that the very definition of renewing a claim? 16:34:47 <flaper87> kgriffs: I tested the regex and it doesn't seem to match, I'll test it a bit further 16:35:01 <kgriffs> does the client need to know/care about the "real age"? 16:35:18 <kgriffs> flaper87: kk, maybe the message changed with a recent Mongo releasse 16:35:30 * kgriffs wishes he didn't have to do a hack that's so fragile 16:35:44 <amitgandhi> #link: https://wiki.openstack.org/wiki/Marconi/specs/api/v1#Update_Claim "The server will reset the age of the claim and apply the new TTL." 16:35:45 <kgriffs> flaper87: we may need to make the regex more flexible 16:35:58 <zyuan_> amitgandhi: yes, i saw that 16:36:04 * flaper87 complained directly with the right guys about those worthless error messages 16:36:36 <kgriffs> so, the simplest fix now is to code to the spec as it now stands 16:36:49 <zyuan_> wait 16:36:50 <zyuan_> wait 16:36:54 <kgriffs> (in my opinion) 16:37:14 <zyuan_> sqlite's age is same as spec 16:37:54 <zyuan_> mongo's is not -_- 16:38:22 <kgriffs> ok, I propose we modify the mongo driver to be consistent with sqlite and the spec 16:38:29 <kgriffs> any objections? 16:38:31 <cppcabrera> +1 kgriffs 16:38:34 <amitgandhi> kgriffs: +` 16:38:36 <amitgandhi> +1 16:38:45 <flaper87> +1 16:38:52 <zyuan_> wait 16:38:52 <Sriram> seems good. 16:39:03 <zyuan_> mongo has some impl issue 16:39:07 <zyuan_> when doing that 16:39:12 <flaper87> feel free to assign that to me! kgriffs 16:39:22 <kgriffs> kk 16:39:29 <zyuan_> the created time is embaded into ObjectId 16:39:46 <kgriffs> flaper87: I think we have an expires field as well, nicht? 16:39:55 <zyuan_> you can either change the ID (i don't you dont); or add an extra field? 16:40:16 <flaper87> kgriffs: right 16:40:23 <cppcabrera> c.e 16:40:28 <cppcabrera> ? 16:40:28 <zyuan_> then... age - expires 16:40:29 <flaper87> we've 2: e and c.e 16:40:29 <kgriffs> kk, I think we are OK then 16:40:34 <flaper87> e is message's expire 16:40:36 <flaper87> c.e is claim's 16:40:53 <zyuan_> ok. if it's possible, add a test 16:40:58 <kgriffs> malini, can you get a repro? 16:41:09 <kgriffs> (and a new test!) 16:41:20 <malini> I cannot add a system test for this https://bugs.launchpad.net/marconi/+bug/1210662 16:41:24 <malini> will be too complex 16:41:30 <malini> We need unit tests for this one 16:41:53 <zyuan_> yes, i mean unit test 16:42:18 <flaper87> kgriffs: ahhh, I knew it, I know what happened 16:42:30 <zyuan_> we don't have to wait for some real time; just post, update, compare age, done 16:42:38 <flaper87> that error was originated by the s/qid/qname/ patch 16:42:39 <kgriffs> ok, let's just make sure there is a test somewhere that first proves it is broken (and pls set to confirmed) then proves it is fixed (when patch is submitted) 16:42:43 <kgriffs> darn 16:42:45 <kgriffs> kk 16:42:52 <flaper87> because the error now has the project in the message 16:42:57 <flaper87> I already created a more general regex 16:43:02 <flaper87> I'll submit a patch soon 16:43:08 <zyuan_> grate 16:43:11 <zyuan_> great 16:43:13 <cppcabrera> +1 flaper87 16:43:14 <zyuan_> -_- 16:43:21 <kgriffs> kk 16:43:28 * cppcabrera watches bugs fixed in real time 16:43:37 <kgriffs> I'll assign those two bugs to Mr. Percoco 16:43:43 <zyuan_> my key stokes alwyas .... errrrahhh 16:43:55 <flaper87> w00000t 16:44:17 <flaper87> kgriffs: question, did this really happen? https://github.com/stackforge/marconi/blob/master/marconi/tests/storage/test_impl_mongodb.py#L52 16:44:25 <flaper87> or that's just a sanity check ? 16:45:55 <kgriffs> flaper87: it must have happened, but I can't remember now for sure - it's been too long 16:46:03 <kgriffs> (and indexes have changed around so much) 16:46:05 <flaper87> kk, no worries 16:46:14 <kgriffs> next bug 16:46:14 <kgriffs> https://bugs.launchpad.net/marconi/+bug/1210633 16:46:40 <kgriffs> I need someone to take this 16:46:42 <cppcabrera> Claim release isn't being honored. 16:46:47 <zyuan_> alej can repo it; i tried with unit test... failed to repo 16:46:50 <zyuan_> (so far) 16:47:01 <cppcabrera> I'll take it. 16:47:04 <kgriffs> sold! 16:47:53 <cppcabrera> I just assigned myself - I'll knock it out after the meeting. 16:47:57 <kgriffs> ok, anything else on the topic of bugs before we move on? 16:48:29 <zyuan_> try sqlite 16:48:35 <zyuan_> if possible 16:48:37 <zyuan_> next time 16:48:46 <cppcabrera> will do, zyuan. 16:48:53 <kgriffs> #topic Triage blueprints 16:48:53 <zyuan_> i guess i implemeted it in a very precise way 16:49:40 <kgriffs> #link https://blueprints.launchpad.net/marconi 16:50:17 <kgriffs> so, go down the line, sorted by series 16:50:23 <kgriffs> perf-testing 16:50:44 <kgriffs> anything we need to update on this (status?) 16:50:44 <flaper87> did you guys take a look at the ones I put here? https://etherpad.openstack.org/queuing-scratch 16:51:00 <kgriffs> ah, that's a good idea 16:51:02 <kgriffs> let's take those first 16:51:20 <kgriffs> https://blueprints.launchpad.net/marconi/+spec/multi-db-support (Should it be deleted as per Monday's call?) 16:51:59 <zyuan_> probably 16:52:02 <flaper87> We said that addressing multi-db wasn't a good plan right now, I think we should close this and re-create it if needed 16:52:07 <kgriffs> based on a recent conversation with 10gen, I think partitioning across several DBs is valuable and would like to see a patch 16:52:13 <cppcabrera> I like the idea of addressing it later. 16:52:28 <kgriffs> oz_akan: thoughts? 16:52:59 <amitgandhi> i kind of want to see the results of the perf testing first 16:53:03 <kgriffs> basically, my take on this is 1. it is valuable, but 2. not sure on the priority 16:53:27 <amitgandhi> i think storage placement takes a higher priority 16:53:40 <flaper87> I think that one is easy to implement 16:53:45 <flaper87> I know oz_akan_ was working on that 16:53:52 <kgriffs> ok, mind if I put it in H3 for now but low priority? 16:54:01 <amitgandhi> +1 16:54:03 <flaper87> +1 16:54:27 <kgriffs> next 16:54:29 <megan_w> seems like the right call 16:54:30 <kgriffs> https://blueprints.launchpad.net/marconi/+spec/use-webob (I think this can be deleted) +1 16:54:44 <kgriffs> and the next one 16:54:51 <kgriffs> I put those in to appease the gods 16:55:23 <zyuan_> .. 16:55:25 <kgriffs> for the greater good, it may be nice to just port marconi over as a science project and see what a difference it makes 16:55:25 <cppcabrera> +1 for deleting pecan/webob bps 16:55:45 <kgriffs> but I'm not super optimistic that it is something that would make it into mainline 16:56:16 <flaper87> lets just close them 16:56:28 <kgriffs> re webob, we can say "we'll take that seriously as soon as swift does" 16:56:39 <flaper87> LOL 16:56:43 <kgriffs> AFAIK we are the only two data APIs so far 16:56:51 <notmyname> :-) 16:56:55 <flaper87> :P 16:56:55 <notmyname> yay data APIs! 16:57:09 <kgriffs> +1 16:57:11 <kgriffs> ok 16:57:16 <Sriram> :D 16:57:26 * notmyname just saw "swift" in here. I don't really have any context 16:57:28 <kgriffs> let me just close those. I think Doug and I have agreed to disagree at this point anyway 16:58:32 <kgriffs> hmm. how do I close a bp? 16:58:39 <zyuan_> lol 16:58:58 <kgriffs> priority "not" ? 16:59:00 <kgriffs> :p 16:59:08 <flaper87> kgriffs: definition 16:59:13 <flaper87> set it obsolete or superseeded 16:59:35 <kgriffs> they need a more accurate term, but oh well 16:59:37 <kgriffs> moving on 16:59:50 <kgriffs> https://blueprints.launchpad.net/marconi/+spec/stream-messages (I think this one can be deleted) +1 16:59:51 <kgriffs> https://blueprints.launchpad.net/marconi/+spec/streaming-serialization (I think this one can be deleted) +1 17:00:03 <kgriffs> I added those in case we wanted to revisit the idea later 17:00:51 <kgriffs> but, if nobody thinks we are going to go back to streaming posted messages, we can forget about it. 17:01:05 <flaper87> I'd close them and re-open them later if needed 17:01:05 <zyuan_> i can remember them 17:01:12 <kgriffs> note that at the moment we don't stream messages when listing them - we serialize to JSON in RAM 17:01:12 <zyuan_> so you can delete them now 17:01:19 <kgriffs> ok 17:01:26 <kgriffs> going once 17:01:30 <kgriffs> going twice 17:01:44 <kgriffs> sold to the guy eating a pop-tart in the back! 17:02:02 <flaper87> LOOOL 17:02:13 <cppcabrera> :P 17:02:36 <kgriffs> https://blueprints.launchpad.net/marconi/+spec/placement-service was superseeded by https://blueprints.launchpad.net/marconi/+spec/storage-placement (Is this for H3 or I-1 ?) (They need to be merged.) 17:02:58 <cppcabrera> If the contents of those two BPs could be merged, it's perfect, IMO. 17:03:15 <kgriffs> ok, let's merge and change the name to "cell" or whatever you want to call it 17:03:16 <flaper87> (the oldest wins) :P 17:03:40 <kgriffs> #action cppcabrera to merge marconi cell bp's 17:04:07 <kgriffs> flaper87: if you think "cell" is going to give people the wrong idea, we are open to suggestions for naming 17:04:16 <kgriffs> moving on 17:04:17 <kgriffs> https://blueprints.launchpad.net/marconi/+spec/routing-api (H-3 or I-1?) 17:04:38 <kgriffs> I think that would be I-1 17:04:44 <kgriffs> what does everyone think? 17:04:57 <kgriffs> I just want to get a solid, baseline release out first 17:05:08 <cppcabrera> I-1, as well as v1.1 or v2.0 of the API 17:05:08 <flaper87> kgriffs: I think cell is going to confuse people, I'd keep storage palcement and leave cell for a higher "cell" concept 17:05:30 <flaper87> I prefer I-1 as well 17:05:33 <kgriffs> flaper87: we've actually been contemplating replacing storage placement with a higher-level cell 17:05:35 <amitgandhi> flaper87: +1 to not using "cell" 17:05:48 <kgriffs> so, routing to webheads + storage as a big black box 17:05:50 <flaper87> kgriffs: mmh, seems you've discussed that service a bit further 17:06:04 <cppcabrera> yup, a bunch flaper87. 17:06:08 <flaper87> when can we sync up on that? (G+?) 17:06:16 <kgriffs> flaper87: yeah, last week, I wanted to discuss as part of this meeting but we are out of time 17:06:28 <kgriffs> let's sync up on all that tomorrow? 17:06:32 <flaper87> kgriffs: +1 17:06:52 <kgriffs> ok, cppcabrera can you schedule somethin' 17:07:24 <kgriffs> #action flaper87, cbbcabrera to schedule another placement service and/or cell architecture discussion 17:07:31 <cppcabrera> Will do. 17:07:34 <flaper87> +1 17:07:38 * cppcabrera is action-packed 17:07:49 <Sriram> woot!:) 17:07:51 <flaper87> cppcabrera: Is the wiki page updated? 17:08:07 <flaper87> I'll read it before our next G+ 17:08:19 <cppcabrera> Yeah, I made sure the latest is up there last I worked on it. It should have lots of haproxy and cell ideas on it. 17:08:31 <cppcabrera> *"cell" 17:08:40 <kgriffs> ok folks, we are out of time. we will take care of the placement service chat in breakout, and we can audit the API next Monday. 17:08:53 <kgriffs> #endmeeting