19:00:09 #startmeeting Poppy Weekly Meeting 19:00:09 Meeting started Thu Jan 8 19:00:09 2015 UTC and is due to finish in 60 minutes. The chair is amitgandhinz. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:00:10 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:00:12 The meeting name has been set to 'poppy_weekly_meeting' 19:00:16 #topic RollCall 19:00:23 welcome back everyone 19:00:27 who do we have today? 19:00:37 o/ 19:01:02 o/ 19:01:14 ..hello ..folks.. 19:01:19 o/ 19:01:31 hi miqui_ 19:01:50 o/ 19:02:07 #link Agenda https://wiki.openstack.org/wiki/Meetings/Poppy 19:02:13 o/ 19:02:38 #topic Last Year This Year on Poppy 19:02:47 #link Last Meeting - http://eavesdrop.openstack.org/meetings/poppy_weekly_meeting/2014/poppy_weekly_meeting.2014-12-18-19.00.html 19:03:14 okay, so action items from last time 19:03:17 o/ 19:03:19 amitgandhinz to set up a design discussion on the Message Queue Driver in early January 19:03:26 ok lets set this up 19:03:58 how does next Thursday sound? 19:04:07 good for me! 19:04:11 maybe at 1pm EST 19:04:12 during usual meeting hrs? 19:04:21 nah lets do it an hour before the usual meeting 19:04:23 I am good. 19:04:35 +1 19:04:40 miqui_: ? 19:04:41 I will be in India & will try to stay awake (A) 19:04:43 +1 19:04:49 mpanetta: ? 19:04:56 Sounds good to me 19:05:15 ok i will send out a calendar invite 19:05:21 miqui_: i dont have your email addr 19:05:42 migmaqer@gmail.com 19:05:48 thanks 19:06:13 the design discussion will center on what drivers we implement, how workers should work, etc 19:07:08 #agreed amitgandhinz will hold a design discussion with the poppy team on the Message Queue Driver implementation on Thursday Jan 15, 2015 at 1pm EST 19:07:23 ok next action item was malini to add doc to cdn vendor wiki page on how to add a mimic plugin for their api 19:07:40 :-$ 19:07:45 can we carry it over? 19:07:51 sure 19:07:55 #action malini to add doc to cdn vendor wiki page on how to add a mimic plugin for their api 19:08:21 lets get this done as I want us to make it really obvious for vendors to build out their drivers now 19:08:37 sure 19:09:00 alright.... 19:09:13 #topic Review Kilo2 Activities 19:09:17 #link https://launchpad.net/poppy/+milestone/kilo-2 19:09:47 so first of all congrats everyone, there was an immense number of bugs fixed over the holidays 19:10:07 yay! 19:10:14 =D 19:10:23 I missed the bug party but still congrats is in order. 19:10:37 we still have some but not many so thats awesome 19:11:19 there is still about 2 weeks left in the kilo 2 cycle 19:11:22 we ate bugs for holidays 19:11:31 yum... 19:11:40 ..hehehe... 19:11:41 so lets target some of these remaining blueprints 19:11:52 especially chocoloate covered 19:12:15 yumm.. 19:12:23 1. Msq Queue Driver - we will discuss this next week and then get started on it 19:12:41 tonytan4ever: Update on shared SSL support? 19:12:58 This is pretty much done, it could use some reviews from everybody though. 19:13:13 so Good progress or In Review? 19:13:13 https://review.openstack.org/#/c/145365/ 19:13:29 Should be In Review 19:13:46 ok. lets get some reviews on this 19:14:14 got it 19:14:15 miqui_: Update Home Doc 19:15:19 malini: Specify provider url in the conf file 19:15:30 hi amitgandhinz, sorry, i have no upda tes..been swamped... 19:15:37 miqui_: thats ok 19:15:38 sorry about htis.. 19:15:47 we have this already in place for akamai & fastly.. 19:16:06 havent started for the rest 19:16:07 its just maxcdn and cloudfront that need updating still 19:16:21 yes 19:16:23 im going to take your name off this for now 19:16:33 thanks! 19:17:30 obulpathi: Create Rackspace DNS plugin for mimic 19:18:03 it still in same state 19:18:17 once I fix the caching rules bug, I will finishes this off 19:18:27 ok 19:18:31 malini: End To End Tests for Poppy 19:18:42 tht is merged 19:18:53 I wud like to make some minor updates 19:19:02 ill mark it as completed 19:19:02 around waiting for DNS propagation 19:19:10 you can submit bugs for the updates 19:19:16 sure 19:20:05 ok so the main items for kilo remaining are the queue driver, ssl support, and fixing the api tests at the gate 19:21:26 does anyone have any questions regarding the bugs? 19:21:47 nope 19:21:52 nope... 19:22:07 no. 19:22:28 #topic New Items 19:22:29 nope 19:22:37 Nothing new on the agenda 19:22:42 #topic Open Discussion 19:22:55 Anyone have something they want to discuss? 19:22:59 amitgandhinz: can we talk a bit abt the conformance tests? 19:23:04 ya 19:23:18 o/ 19:23:34 malini: first, then megan_w_ 19:23:39 if anyone from the edgecast/verizon team is here, i'd love to get feedback on how their plugin is coming along 19:23:41 sounds good 19:24:02 * amitgandhinz watches megan_w_sneak her question in haha 19:24:13 amitgandhinz: if we were to write a conformance test for an existing provider, say akamai or fastly - what is the o/p we want to compare against in the conformance test? 19:24:41 so we want to make sure they implement the interface correctly and can handle the different inputs 19:24:51 so all providers have to meet the abstract interface we defined 19:25:10 you mean this one https://github.com/stackforge/poppy/blob/master/poppy/provider/base/services.py ? 19:25:12 i want the conformance tests to be able to pass a range of different values (postive/neg tests) to those interfaces 19:25:15 ya 19:25:30 and the conformance tests should assert the correct responses are returned 19:25:44 that way as a provider builds their driver, they know that they are meeting the contracts 19:26:02 ok..I'll start working on a POC & solicit feedback as I go 19:26:07 I am adding a bp for tht now 19:26:21 to answer the question on if the provider driver configures the cdn settings correctly at their end, the api tests and end to ends tests should catch that 19:26:42 we will need a way to run the api tests multiple times, once for each provider we have 19:27:04 we can just create a big happy flavor with all providers in it :) 19:27:08 that way we can see which providers are working and which ones could be broken 19:27:46 we should be able to just use different test.conf files right? to test each provider flavor on their own 19:28:07 yeap..tht will work too 19:28:52 that way we can setup each conf file with its appropriate provider credentials, and also point it to use mimic if necessary 19:29:25 sounds good.. 19:29:30 megan_w_: your turn 19:29:49 fantastic. is there anyone from the edgecast/verizon team here? 19:29:56 (could be a quick turn) 19:30:14 wait for it... 19:30:22 ... 19:30:35 ok, i'm good. nothing for me :) 19:31:04 shud have let megan_w_ go first :) 19:31:09 can i ask a question.... 19:31:32 * miqui_ raises hand... 19:31:39 yes please 19:32:00 go for it miqui_ 19:32:13 for the message queue driver design discussion is there a particular messaing backend to consider? 19:32:26 i want to support the following: 19:32:33 just asking so as to t"think ahead " somewhat... 19:32:45 1. oslo.message - this gives us rabbitmq, zeromq, and qpid, and maybe kafka support 19:32:56 2. zaqar support 19:33:09 3. maybe celery or taskflow 19:33:19 but 1 and 2 are required 19:33:28 k, thanks... 19:33:52 if we build the driver generic enough we should be able to support many queue platforms 19:34:04 right + 19:34:08 +1 19:34:13 but 1 and 2 tie us into the openstack platform nicely 19:34:49 ok any more questions? 19:35:12 am good for now... thanks... 19:35:17 do we want to start talking a bit more abt queue driver now? 19:35:24 since we have 30 minutes left? 19:35:48 lets do some prep first 19:35:48 I don't want any life back :D 19:35:57 hah! 19:35:59 ok 19:36:04 i want to gain a better understanding myself of oslo.messaging 19:36:13 sure 19:36:17 i think i know zaqar pretty well (one hopes =p) 19:36:48 and i know mpanetta has some concerns around worker processes that need to be heard 19:36:58 yeah 19:37:12 mainly that they die properly :P 19:37:26 so lets spend the week building up our notes/concerns, and then we can hit hard at the thur meeting 19:37:37 we can also talk about it during the week too =) 19:37:42 i have some experience with rabbitmq 19:37:51 cool 19:38:12 regarding the worker mechanisms 19:38:14 ..but likewise need to ramp up on oslo.messaging... 19:38:57 Celery use rabbitmq as a backend queue system, and it has a nice GUI interface to manage and monitor all the worker processes. 19:39:01 I am just saying. 19:39:18 if we have threading based workers (like greenlets), poppy will be able to handle much higher load (compared to process based) 19:39:45 a significant portion of time, the workers spend in waiting for network io state 19:40:10 this limits the number of messages that can get processed 19:40:23 My main concern is that we can mange the processes with something like supervisord or circus. 19:40:41 if we use process based workers, we can not have large number of workers (swapping, memory etc ..) 19:40:42 worker pool? 19:41:15 for every network io, provider call , DNS call, they need to be swapped, which created lot of overhead 19:41:33 it amy be worth looking into how designate is doing their workers (just for comparison) 19:41:38 s/amy/may 19:41:41 good idea 19:42:04 i like how nodejs model... 19:42:07 does any other poppy project use queue/worker mechanism? 19:42:18 how it handles multiple load use cases.. 19:42:34 not sure if any of you have read some about this javascript v8 engine... 19:42:44 It is all event driven isn't it? 19:42:51 yes 19:42:57 it is very very fast,,,, 19:43:06 miqui_: unfortunately nodejs isnt openstacky =/ 19:43:13 bu i think we have stick to python 19:43:17 but there are non blocking python libs 19:43:22 yup amitgandhinz...you are right... 19:43:23 twisted 19:43:27 like twisted 19:43:29 =P 19:43:29 asyncio 19:43:30 etc 19:43:38 asyncio is py3+ 19:43:43 * mpanetta is probably still on an ansync kick... 19:43:57 Twisted is kinda cool, but the docs are kinda lacking. 19:43:57 nsync was so 90s ! 19:44:05 haha 19:44:14 s/ansync/async/ :P 19:44:39 im not sure if twisted is acceptable .... 19:44:49 ..hehehe 19:44:52 so lets research this this week also 19:44:59 come up with ideas for thursday 19:45:20 #action investigate how to do worker processes with non blocking io 19:45:28 #action research oslo.messaging 19:45:40 Are there any openstack preferred libs for this? 19:45:51 idk 19:46:20 got it 19:46:59 ok.... 19:47:03 any thing else ? 19:48:05 nothing from me 19:48:13 none from me too.. 19:48:19 none from me... 19:48:24 One thing from me. 19:48:53 go for it tonytan4ever 19:49:15 A quick FYI, the domain field in Poppy service will have a "shared ssl" field, to support shared ssl domain 19:49:25 `shared_ssl` - optional: whether to use shared ssl certificate on this domain, right now only support true. Defaults to false. 19:49:45 This is what I will put on apiary doc. 19:50:18 right. its also worth noting that if shared_ssl = true, then the domain entered will be overwritten with a generated domain using the shared cert 19:50:46 overwritten in the service details? 19:50:47 Yeah. 19:50:52 yeah 19:51:05 because a custom user domain doesnt make sense with a shared domain 19:51:17 you cant cname your own domain to a shared domain with SSL 19:51:25 the cert would be rejected in the browser 19:51:28 in tht case, does it make sense to still keep domain a required field? 19:51:43 You are right, it does not. 19:51:48 thinking about making it not required if shared_ssl = true 19:51:53 +1 19:52:51 cool =) 19:53:02 anyone else? 19:53:14 i think everyone has asked a question =P 19:54:39 guess tht is it? 19:54:45 I think so 19:54:48 1 min never seemed longer :D 19:55:11 hehe 19:55:15 ok thanks everyone 19:55:17 Sooo quiet 19:55:19 #endmeeting