08:03:40 <joehuang> #startmeeting multisite
08:03:40 <collabot`> Meeting started Thu Mar 10 08:03:40 2016 UTC.  The chair is joehuang. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:03:40 <collabot`> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:03:40 <collabot`> The meeting name has been set to 'multisite'
08:03:52 <joehuang> #topic rollcall
08:03:52 <sorantis> #info dimitri
08:03:57 <joehuang> #info joehuang
08:03:59 <zhipeng> #info zhipeng
08:04:01 <Malla> #info Malla
08:04:20 <SAshish> #info Ashish
08:04:30 <joehuang> #topic release
08:05:09 <joehuang> Dimitri brings a topic about release, for OpenStack Mitaka is approching
08:05:11 <sorantis> so, we’ve got a couple of commits about to land on master
08:05:37 <joehuang> shall we follow the tempo or not, let
08:05:39 <sorantis> if everything goes well, this will give us an end-to-end quota management for nova
08:05:49 <sorantis> now,s
08:06:01 <SAshish> yes
08:06:05 <joehuang> please
08:06:19 <sorantis> since we have all our CI and repos set up in an OpenStack manner I think we should also follow the openstack pace
08:06:45 <sorantis> this doesn’t mean that we skip OPNFV releases
08:07:03 <sorantis> actually we will have to ensure that our integration of KB also works in each release of OPNFV
08:07:12 <sorantis> so we’ve got plenty to do for C as well
08:07:33 <sorantis> but I consider this as a separate activity not directly related to KB development as such
08:07:53 <sorantis> I would like to hear your thoughts
08:08:20 <zhipeng> release engineering generally is a tiresome work
08:08:44 <zhipeng> for official projects, they have the assisstance from openstack-infra team
08:08:49 <joehuang> From my point of view, how much engineering works to be done before the release. This is official release requirements https://github.com/openstack/releases
08:08:54 <zhipeng> in opnfv we have our infra team
08:09:16 <zhipeng> but unofficial projects we could only do it ourselves
08:10:15 <sorantis> how big is the release work?
08:10:29 <joehuang> a todo list for a release is needed, we may create an ethepad to track these tasks
08:10:56 <joehuang> if only tagging, it's much more simpler
08:11:29 <joehuang> but generally we need to provide documentation, installation package, release notes
08:11:42 <sorantis> I think for Mitaka we can have a simple tagging
08:12:09 <sorantis> since we’re not an official OS project, we can skip the requirements, as nobody will follow up on those in OpenStack anyways
08:12:19 <sorantis> and we prepare all this in OPNFV
08:12:29 <zhipeng> and also we need define test scenarios to at least verify it could work wtih Nova
08:12:52 <sorantis> yes, there will be FTs for that
08:13:15 <sorantis> so I hear that you guys want to keep the pace of OPNFV
08:13:47 <sorantis> what’s your plans with Tricircle?
08:13:50 <zhipeng> OPNFV is a hard deadline
08:13:59 <joehuang> we have no ST
08:14:10 <zhipeng> first release with OPNFV C
08:14:30 <zhipeng> what I suggest is that no hard deadline set for Mitaka
08:14:35 <zhipeng> do it as best effort
08:14:47 <zhipeng> if it works then hurray, if not no biggy
08:14:48 <joehuang> After finishing basic functionalities, tricircle will launch a release, but it still take times for the first release
08:15:56 <SAshish> looks like Dimitri has lost connection
08:15:57 <zhipeng> sorantis quits in fury lol
08:15:58 <joehuang> After the first release, then follow the pace of openstack, to release every 6 month
08:17:03 <joehuang> Dimitri's connection is lost
08:17:14 <SAshish> He will join in a moment
08:17:33 <joehuang> Hi, Ashish, what about your idea
08:17:39 <zhipeng> stormed out of the irc door lol
08:18:29 <sorantis> sorry, the internet died
08:18:33 <sorantis> could you resend?
08:19:01 <SAshish> OPNFV is a hard deadline
08:19:05 <SAshish> first release with OPNFV C
08:19:11 <SAshish> <zhipeng> what I suggest is that no hard deadline set for Mitaka
08:19:19 <SAshish> <zhipeng> do it as best effort
08:19:19 <joehuang> For tricircle, the first release will not follow OpenStack pace, after that will release at the same time with OpenStack every 6 month
08:19:33 <sorantis> ok, fair enough
08:19:44 <sorantis> but let’s at least have tagging
08:20:05 <joehuang> I think the biggest challenge is installation/configuration work in OPNFV environment
08:20:22 <joehuang> to release a version is not un-acceptable task
08:20:22 <zhipeng> sorantis that is doable
08:20:43 <sorantis> good
08:20:47 <SAshish> +1
08:21:11 <sorantis> I agree. we need to focus on integration in OPNFV after we will have complete the commits
08:21:18 <sorantis> so let’s do like so
08:21:34 <sorantis> 1. Finilize work in KB for the first mile-stone
08:21:44 <sorantis> 2. Tag for mitaka
08:21:53 <joehuang> we also need tempest or like integration test before release
08:21:54 <sorantis> 3. Work on integration in OPNFV
08:22:19 <joehuang> good summary
08:22:20 <sorantis> 1.5 Create FTs for KB
08:22:36 <zhipeng> yes , we need those for OPNFV release anyways :)
08:22:38 <joehuang> time for the first milestone
08:23:05 <SAshish> for FTs we will need clients
08:23:18 <SAshish> kbclient
08:23:25 <sorantis> or curl
08:23:31 <joehuang> or use CURL for the first release
08:23:38 <SAshish> so for now we will have python restclients or curl
08:23:56 <joehuang> we have no enough time frame to develop client
08:24:03 <sorantis> +1
08:24:04 <SAshish> yes
08:24:07 <SAshish> agree
08:24:22 <sorantis> i think we’re on the same page with just using curl for the first FT suite
08:24:23 <joehuang> use curl to create a basic FT test cases set
08:24:34 <joehuang> +1
08:24:48 <SAshish> yes Dimitri, As we discussed.
08:24:54 <joehuang> and FT test environment setup
08:25:08 <joehuang> need at least two regions
08:25:38 <SAshish> devstack based?
08:25:48 <joehuang> the devstack cannot support two region yet, we have to manually fix the bug in devstack scripts
08:26:15 <joehuang> or kolla based?
08:26:38 <joehuang> If we have installation by kolla.
08:27:15 <SAshish> which ever is simple
08:27:57 <sorantis> I’ve got no strong opinion about this
08:27:59 <joehuang> so my point of view is that we need to prepare basic multiregion environment in opnfv testbed, so that we can run this, otherwise a lot of work to do
08:28:00 <sorantis> whichever works
08:28:18 <sorantis> so this work should be in OPNFV
08:28:36 <sorantis> and I guess the CI team will help us out with it? :)
08:28:45 <joehuang> for FT test, by devstack, or by OPNFV testbed
08:29:52 <sorantis> ultimately we need both
08:30:05 <SAshish> so guys we have 1, 1.5 & 2  for mitata?
08:30:22 <sorantis> yes
08:30:33 <sorantis> 1 and 2 should be doable within this week
08:30:39 <sorantis> the commits are ready
08:30:45 <sorantis> except one is failing at the moment
08:30:54 <sorantis> but your comments have been addressed joe
08:31:15 <SAshish> there is merge conflicts
08:31:19 <sorantis> then we should address 1.5
08:31:43 <SAshish> yes. will resolve
08:31:45 <sorantis> since we’ve agreed to do a simple tagging, it means we have time
08:32:28 <joehuang> to Dimitri, use this API calling for usage retrieve: http://developer.openstack.org/api-ref-compute-v2.1.html#listDetailQuotas
08:32:34 <SAshish> Joe, shall we spend some time here to discuss you?
08:32:42 <joehuang> ok
08:32:43 <SAshish> to discuss your comments?
08:32:54 <joehuang> just paste a link above
08:33:03 <joehuang> for the usage retrievement
08:34:39 <SAshish> okay.. do you know the corresponding command for the same?
08:34:46 <SAshish> I can try that out now
08:35:02 <joehuang> no command, but url
08:35:18 <joehuang> or use nova limits
08:35:57 <joehuang> a little different between nova limits and listDetailQuotas
08:36:05 <joehuang> the response is different
08:37:12 <SAshish> root@admin-mon-node1-osd0:~/kingbird# curl -g -i -X GET http://172.16.73.126:8774/v2.1/b7efb082377042c084ac145d1ad82200/os-quota-sets/b7efb082377042c084ac145d1ad82200/detail -H "User-Agent: python-novaclient" -H "Accept: application/json" -H "X-OpenStack-Nova-API-Version: 2.12" -H "X-Auth-Token: $TOKEN" HTTP/1.1 200 OK Content-Type: application/json Content-Length: 919 X-Openstack-Nova-Api-Version: 2.12 Vary: X-OpenStack-Nova-A
08:37:25 <SAshish> {"quota_set": {"injected_file_content_bytes": {"reserved": 0, "limit": 10240, "in_use": 0}, "metadata_items": {"reserved": 0, "limit": 100, "in_use": 0}, "server_group_members": {"reserved": 0, "limit": 10, "in_use": 0}, "server_groups": {"reserved": 0, "limit": 10, "in_use": 0}, "ram": {"reserved": 0, "limit": 50176, "in_use": 2560}, "floating_ips": {"reserved": 0, "limit": 10, "in_use": 0}, "key_pairs": {"reserved": 0, "limit
08:37:43 <joehuang> yes, detail usage is returned
08:38:32 <SAshish> >>> aa['quota_set']['metadata_items'] {'limit': 100, 'reserved': 0, 'in_use': 0} >>> aa['quota_set']['ram'] {'limit': 50176, 'reserved': 0, 'in_use': 2560} >>>
08:39:16 <SAshish> yes. so it is for a project in one region
08:39:24 <joehuang> correct
08:39:36 <SAshish> strangely, we dnt have any API call for this
08:40:48 <SAshish> Am I right?
08:40:56 <SAshish> No API for this?
08:41:06 <joehuang> client doesn't support it, but curl support it
08:41:33 <joehuang> or you can try limits
08:41:38 <joehuang> nova limits
08:42:07 <joehuang> summary of tasks, please check to see if it's correct
08:42:24 <joehuang> #1. Finilize work in KB for the first mile-stone with OpenStack mitaka
08:42:36 <joehuang> #info 1. Finilize work in KB for the first mile-stone with OpenStack mitaka
08:42:45 <SAshish> yes. nova limits
08:42:49 <SAshish> thanks
08:42:57 <joehuang> #info 1.5 CURL FT test
08:43:05 <joehuang> #info 2. Tag for mitaka
08:43:13 <joehuang> #info 3. Work on integration in OPNFV
08:43:49 <joehuang> nova limits use list detail like information, but combine the reserve and in-use
08:44:05 <joehuang> and the presentation name is different, max_xxxx
08:44:33 <SAshish> yes. will use nova limits
08:44:49 <SAshish> there will be API for this
08:44:59 <joehuang> how about the pypi /tarball for a release?
08:45:59 <zhipeng> that would be ideal
08:46:17 <joehuang> novaclient support nova limits, but not support "list detail"
08:46:17 <zhipeng> easy to integration easy to test
08:46:18 <SAshish> yes
08:46:54 <joehuang> I don't know how much work needed for pypi
08:47:25 <zhipeng> for a scale like KB and TC, would need some work
08:47:42 <zhipeng> we delivered pypi for Parser, but Parser is really small project
08:48:17 <joehuang> ok, this could be the 4th one
08:48:29 <SAshish> fine
08:48:38 <SAshish> can all projects be finished for the sync. before the next timer task comes?  What will happen if two timer tasks running overlapping?
08:48:58 <SAshish> we will keep time interval as big as 20 mins
08:49:00 <joehuang> #info 4. preparation of pypi/tarball for a release
08:49:15 <sorantis> +1
08:49:23 <SAshish> for?
08:49:58 <joehuang> the 4 could be done after the 3 is ready
08:50:08 <sorantis> kingbird
08:51:38 <joehuang> Zhipeng, do you know CI / Installation use tarball or pypi to install an OpenStack service?
08:52:03 <zhipeng> im not familiar with that
08:52:04 <joehuang> or other package format?
08:52:25 <zhipeng> in OPNFV we just made the pypi package avaibale online
08:52:37 <zhipeng> and then testing and integration team pull that down and do their work
08:52:45 <sorantis> pypi will anyway be needed as openstack services install dependencies using pip
08:52:52 <zhipeng> yes
08:53:33 <joehuang> ok, so pypi is necessary
08:54:37 <joehuang> back to the review comment, project number is not fixed, so 20 minutes may not be longer enough for all project quota's sync.
08:58:13 <joehuang> we may have one loop task, once all finished, then retrieve projects list again, and do every batch until all finished, and then next cycle
08:58:26 <SAshish> suppose we have 100 projects and it starts batch from begening, projects in a batch come come in series
08:59:23 <joehuang> if we use a timer, then the issue is just what mentioned, maybe not finished, but next timer comes
08:59:51 <joehuang> maybe a loop for the last timer has not finished, but next timer comes
09:00:33 <SAshish> yeah. chances are there.
09:01:22 <joehuang> how do you think about my suggestion " one loop task, once all finished, then retrieve projects list again, and do every batch until all finished, and then next cycle"
09:02:18 <joehuang> We may discuss this in M-L.
09:03:12 <sorantis> yes
09:03:13 <SAshish> yes Its needed
09:03:27 <joehuang> ok thanks for the meeting
09:03:31 <sorantis> please finalize the discussion in ML
09:03:33 <sorantis> thanks
09:03:58 <joehuang> #endmeeting