19:00:16 #startmeeting Poppy Weekly Meeting 19:00:16 Meeting started Thu Sep 4 19:00:16 2014 UTC and is due to finish in 60 minutes. The chair is amitgandhinz. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:00:17 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:00:19 The meeting name has been set to 'poppy_weekly_meeting' 19:00:38 #topic rollcall 19:00:42 o/ 19:00:47 hi Poppy folks! 19:00:53 who do we have here for the poppy meeting today? 19:00:59 <- 19:01:11 Yo. 19:01:19 o/ 19:01:20 hello 19:01:53 o/ 19:02:09 hi obulpathi 19:02:14 long time no see 19:02:19 you are so fast! 19:02:30 oz_akan: ^ 19:02:38 o/ 19:03:32 obulpathi: should I get the blue pill? 19:03:36 anyone from fastly/maxcdn/akamai today? 19:04:42 oz_akan: sorry, I already took the red pill. 19:04:54 mikehn: since you are new here, could you give another quick intro like you did on the #openstack-poppy channel 19:05:17 oz_akan, obulpathi: I know this is the poppy meeting, but can you stop talking pills? 19:05:29 hello mikehn! 19:05:42 ) 19:05:44 :) 19:06:32 Hello mikehn, it's a pleasure to see you here. 19:06:55 Another new guy to #openstack-poppy channel. 19:07:02 hello nitin__! 19:07:06 ah yes nitin__, welcome! 19:07:07 Hello, 19:07:14 can we get a quick intro from you too 19:07:14 thanks ! 19:07:20 Sure 19:07:26 hello there nitin__ 19:07:50 Nitin Agarwal , Software Developer and Programmer pursuing B.tech (Computer science) 19:08:13 I had spent my Summer 2014 at CERN, European Organisation for Nuclear Research. 19:08:48 welcome =) 19:08:49 nitin__, where do you study? 19:08:52 glad to have you here nitin__! 19:08:55 Evaluated Orchestrating Docker on OpenStack. 19:09:17 obulpathi: International Institute of Information Technology Hyderabad, INDIA. 19:09:52 nitin__: cool, welcome to Poppy 19:09:56 cool. welcome to the team nitin__ 19:09:56 amitgandhinz: sure 19:10:26 at cern, http://openlab.web.cern.ch/about/people/nitin-agarwal 19:10:38 Would love to visit CERN one day. 19:10:49 thanks amitgandhinz obulpathi 19:11:13 welcome nitin__ 19:11:15 <- Michiel, product management at Hibernia Networks, responsible for ip,ethernet & cdn 19:11:21 oz_akan: thanks ! 19:11:38 thanks mikehn 19:11:54 ok, lets move on, a few topics to get through today 19:12:10 #topic: Review last weeks action items 19:12:19 #link: http://eavesdrop.openstack.org/meetings/poppy_weekly_meeting/2014/poppy_weekly_meeting.2014-08-28-19.00.html 19:12:42 ok, first one is me. I still have not reviewed the MaxCDN CDN Manager API. 19:12:57 I will roll it over to next time again. 19:13:14 #action amitgandhinz to investigate MaxCDN CDN Manager API for master/sub accounts 19:13:47 ok next item was to dig more into the account management system that was being discussed with fastly 19:14:06 no progress has been made here either. I will probably wait for megan_w to return to discuss that item 19:14:24 #action megan_w, guimarin, edward-fastly, and amitgandhinz to look into what "building out an account management system" is about 19:14:49 ok last item, obulpathi to keep bugging the atlanta openstack meetup organizers 19:14:55 any update there? 19:15:01 no one has got back to me so far 19:15:05 I emailed them again 19:15:14 ok 19:15:22 #action obulpathi to keep bugging the atlanta openstack meetup organizers to get on the schedule 19:15:22 I will discuss about the poppy talk in this months meetup 19:15:33 sounds good 19:15:35 and get it scheduled 19:15:52 #topic Updates on Blueprints and Bugs 19:16:05 #link https://blueprints.launchpad.net/poppy 19:16:38 #link: https://blueprints.launchpad.net/poppy/+spec/create-service 19:16:43 tonytan4ever: any updates on this? 19:17:14 Working on it, my evaluation is about 30% done/ 19:17:43 Should be able to get the PR out for review around next Monday. 19:17:49 ok 19:18:05 lets try to get it out incrementally - maybe the transport/manager/storage parts 19:18:12 followed by each provider integration 19:18:27 it might make the patchsets smaller that way and faster feedback and turnaround 19:18:41 Sounds good. 19:18:46 ok next one... 19:18:50 #link https://blueprints.launchpad.net/poppy/+spec/flavors 19:18:54 this one is mine 19:18:58 desperately needs reviews 19:19:16 I am reviewing this pr 19:19:24 I am reviewing and feedbacking, no worries. 19:19:29 its a little bit, so taking time 19:19:30 me too 19:19:33 #link: https://review.openstack.org/#/c/115130/ 19:19:44 yeh sorry, its a biggie 19:19:48 lots of layers 19:20:38 And this one has some overlap with my next link poppy-models as well. 19:20:55 nice segway 19:20:57 #link https://blueprints.launchpad.net/poppy/+spec/poppy-models 19:21:12 the status still shows started 19:21:19 this is making good progress right? 19:21:25 Yes 19:21:34 Actual status is in-review. 19:22:08 I should be able to roll out the latest PR with review comments addressed by today. 19:22:13 ok cool 19:22:19 lets try to make this one land 19:22:55 ok i will discuss the next bp regarding the sqla driver later (have an agenda item for it) 19:23:12 #link https://blueprints.launchpad.net/poppy/+spec/health 19:23:16 obulpathi: updates? 19:23:41 Its mostly done. I am stuck at routing urls dynamically in pecan 19:24:06 ocne I fix it, will do a final pr 19:24:11 ok cool 19:24:17 i'll keep it at Good Progress then 19:24:23 yes, thanks 19:24:44 ok lets go to the next topic 19:24:52 #topic Limit Counts 19:25:08 What limit counts should exist on adding origins/backends for a service/zone/distribution. 19:25:36 so i think we determined that fastly had no limits for backends that was apparent from their docs 19:25:50 max cdn had just one origin per zone??? 19:26:25 aaron-Ma1CDN: ^^ 19:26:27 link# http://docs.aws.amazon.com/general/latest/gr/aws_service_limits.html#limits_cloudfront 19:26:42 you can find the CloudFront limits in the above link 19:27:08 cloudfront has a limit of 25 per distribution 19:27:28 mikehn: im not familiar with hybernia cdn 19:27:40 does it limit the number of origins in a service definition? 19:28:37 let me look how you define 'service' 19:28:50 similar to cloudfronts distributions 19:28:55 maxcdn;s pullzones 19:28:59 fastly's service 19:29:41 a service is what a customer creates when he wants to serve his site through a cdn provider 19:29:51 we have a staged structure which is account / username / site 19:30:07 then its the site 19:30:18 and for 'site' (which is service?) there is no limit of # of site's configured under a tuple of {account, username} 19:30:46 for a single site, do you allow multiple origins? 19:31:01 at this moment not 19:31:17 ok. thanks 19:31:37 mikehn: do you have any API docs et. which we can look at? 19:31:39 mikehn: do you have a link to the rest api docs for hyberniacdn? 19:31:41 yes 19:31:44 malini: jinx haha 19:31:51 https://portal.hiberniacdn.com/api/ 19:32:26 got a number of references back to our global order entry portal, so skip that :) 19:33:15 ok so i think the conclusion is that the number of supported origins differ widely between providers 19:33:24 yep 19:33:32 either we limit support to 1 origin per site/service/distribution/pullzone 19:33:39 or we let the operator define the min/max 19:33:52 if you want to abstract it, i vote for the latter 19:33:52 or we let the operator define it per flavor 19:33:58 it would be a shame for poppy to limit to 1 19:34:05 +1 to that 19:34:11 +1 19:34:19 i agree, and it depends on what providers an operator chooses to offer 19:34:39 leaving it to the operator to define it sounds good 19:34:44 if we define it per flavor, then the operator has to do homework on the mins supported 19:34:52 min = always 1? 19:35:00 without origin, it's kinda hard to serve content 19:35:12 :) 19:35:14 yes 19:35:18 sorry i think i meant max lol 19:35:25 =P 19:35:58 I think we can define a min=1, and let operator define max 19:36:09 ok, so we make it as part of the flavor definitions the operator creates 19:36:42 #link: http://docs.cloudcdn.apiary.io/#flavors 19:37:13 amitgandhinz: you mean max value will be specified as part of create flavor API call? 19:37:13 any more questions on this? 19:37:18 yes 19:37:26 and it would exist per flavor (not per provider) 19:37:39 But we dont have patch on flavors yet 19:37:49 So updating tht value will be a pain 19:38:02 send the whole blob 19:38:03 why cant it be just a configurable value? 19:38:26 because its flavor based, and operators (eg oz_akan) wants an api to manage it 19:38:31 similar to nova flavor management 19:38:50 conf values per flavor are stored in the datastore 19:38:57 ok 19:39:13 we can discuss implementation more outside this meeting 19:39:19 sure 19:39:26 but i think we agree on the approach 19:39:29 +1 19:39:48 #agreed max limits to be defined with the flavors by the operator 19:40:02 ok moving on now... 19:40:15 #topic Acceptable TTL values and Granularity 19:40:41 the min ttl setting on assets vary greatly between providers 19:41:07 i saw some starting at a min TTL of 5 minutes, others at 1 hour etc 19:41:22 and then the increments also differ in what is supported 19:41:31 how should we handle this variety? 19:42:26 again, let the operator define it 19:43:01 Again, per flavor ? 19:43:01 +1 19:43:20 yeah 19:43:21 so operator defines min and max ttl 19:43:26 what about increments 19:43:56 let operator define it 19:44:00 let the operator define the factor of again 19:44:12 hmmm…with increments it really sounds like we'll need to specify excatly what those values could be 19:44:31 since it can't be really a factor 19:44:37 if we say the range is {1, 6000) and increment is 5 .... 19:44:38 1 hr vs 1 month vs 6 months 19:44:40 malini: and there you have the first difficulty when you want to aggregate and abstract the 100's of cdn-api's out there 19:44:50 mikehn: :) 19:44:54 then we could do 1, 5, 10 .... ? 19:45:10 amitgandhinz: tht still wont be good enough 19:45:19 we always represent ttl and increments in minutes 19:45:27 or should we just round the user supplied value to the nearest supported by the flavor? 19:45:31 so either define it fixed (1,5,10) and deal with the consequences, or make sure you have a variable input & output 19:45:33 since providers could allow 1 & 10, but not 5 19:45:33 so if you want one month = 30 days = n minutes 19:45:59 amitgandhinz: or make it like 1d, 1w, 5h, 5s 19:46:02 malini: if providers support 1 and 10, then increment would be 10 19:46:09 then you dont have that discussion 19:46:14 mikehn: hmmm.... 19:46:23 and we would convert the metric from a human readable format into a machine one 19:46:36 amitgandhinz: providers dont always need to allow increments. 19:47:01 malini: indeed, so keep the option open for "not defining it" 19:47:54 ok i like that. makes it really flexible 19:48:00 default null, otherwise 19:48:13 or if we do want increments, then increment becomes the smallest unit possible, eg 1m 19:48:18 s/do/dont 19:48:51 amitgandhinz: provider's api is definitely checking those values as well, so keep in mind where you want to do conformity checks 19:49:32 yeh, when we send the configs on to the providers, its always possible for them to reject it for whatever reason. 19:49:42 that comes back in the status response 19:49:55 amitgandhinz: good, so my suggestion is to not vote for your "or if we do want increments, then increment becomes the smallest unit possible, eg 1m" 19:50:23 and move that portion to the interaction with the provider 19:50:55 we should let the operators define allowed values for ttl, if none specified allow all 19:51:18 ok, so i think we are all leaning towards operator defining the min/max ttl for a flavor, and the increments (null, 1m, 1h, 1d, 1w, etc) 19:51:37 +1 19:51:38 where 1 is really n 19:51:44 yes 19:51:53 good, next topic ;p 19:51:59 #agreed operator defining the min/max ttl for a flavor, and the increments (null, 1m, 1h, 1d, 1w, etc) 19:52:16 #topic deprecate mockdb driver in favor of sqla driver 19:52:23 ok this one should be quick 19:52:37 the mockdb driver feels pointless and a pain to maintain and serves no purpose 19:52:46 mocking is done using unittest mock 19:52:46 19:53:02 so lets to sqlalchemy which adds value to those not needing cassandra 19:53:11 amitgandhinz: +1 19:53:15 +1 19:53:20 Make sense for me. 19:53:27 #agreed: eprecate mockdb driver in favor of sqla driver 19:53:32 easy 19:53:40 #topic Open Discussion 19:53:49 ok we have 7 minutes in here 19:54:09 any other topics ppl want to discuss before we move back to the #openstack-poppy channel? 19:54:18 do the new folks need any help getting started etc.? 19:54:27 devkulkarni: are you here? 19:54:44 malini: let's do that in #poppy 19:54:55 malini: yes 19:55:22 are #openstack-poppy and #poppy different channels . 19:55:22 We are always on #openstack-poppy. Feel free to ping any of us there 19:55:30 Sure 19:55:32 ok any other topics to discuss here? 19:55:32 nitin__: welcome to poppy once again. 19:55:44 obulpathi: thanks again. 19:55:58 ok thx for the meeting 19:56:09 thanks for attending mikehn 19:56:15 and nitin__ 19:56:18 thanks for participating mikehn 19:56:33 ok lets carry on chatting in the poppy channel 19:56:33 no biggie 19:56:39 thanks everyone 19:56:46 #endmeeting