08:04:21 <joehuang123> #startmeeting multisite
08:04:21 <collabot`> Meeting started Thu Mar 17 08:04:21 2016 UTC.  The chair is joehuang123. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:04:21 <collabot`> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:04:21 <collabot`> The meeting name has been set to 'multisite'
08:04:25 <SAshish> Hello All
08:04:33 <joehuang123> hi
08:04:52 <joehuang123> Mitaka will be released on Apr.8
08:05:21 <joehuang123> #topic rollcall
08:05:26 <joehuang123> #info joehuang
08:05:36 <SAshish> #info SAshsih
08:05:41 <sorantis> #info dimitri
08:06:15 <joehuang123> So I think it would be ok to have a release 0.1 for Kingbird, that means not full release
08:07:15 <joehuang123> last meeting : 1. Finilize work in KB for the first mile-stone with OpenStack mitaka
08:07:18 <SAshish> yes. If it can then, what all we need for this release
08:07:28 <joehuang123> 1.5 CURL FT test
08:07:31 <sorantis> with a tag 0.1 or release mitaka?
08:07:43 <joehuang123> 2. Tag for mitaka
08:08:31 <joehuang123> we can use the tag for Mitaka, but the release number maybe 0.1.0? not the formal 1.0.0
08:08:43 <sorantis> I’m fine with that
08:08:51 <SAshish> fine
08:08:53 <sorantis> as long as we have a version
08:09:03 <joehuang123> yes
08:09:52 <joehuang123> The challenge here is to prepare the FT test environment
08:10:12 <joehuang123> If we only using one devstack node, it'll be easy
08:10:24 <SAshish> we are almost at verge of finishing 1.
08:10:27 <SAshish> so 1.5 is next
08:10:45 <SAshish> just one more commit then we are done with 1
08:10:59 <SAshish> the locking one
08:11:14 <joehuang123> but if we want to use two nodes FT, it's more difficult
08:11:25 <joehuang123> Yes.
08:11:36 <joehuang123> Just one comment on the locking
08:11:45 <SAshish> a fuel based setup
08:11:58 <SAshish> single node
08:12:02 <SAshish> with multi regions
08:12:11 <joehuang123> Can fuel setup multiple regions?
08:12:16 <SAshish> yes
08:12:21 <sorantis> multiple environments
08:12:32 <sorantis> with separate keystones
08:12:33 <SAshish> yeah
08:12:45 <joehuang123> I am not family with fuel
08:12:46 <sorantis> so manual work to reconfigure auth will still be required
08:13:10 <joehuang123> configure the second one to use the node1 keystone
08:13:22 <sorantis> yes, but manually
08:13:23 <SAshish> yes, also for multiple controllers, there will be internal IPs, so kingbird also needs to sit on same node
08:13:26 <joehuang123> each service configuration file need to be modified
08:13:33 <sorantis> correct
08:13:47 <sorantis> it can be scripted of course
08:14:00 <SAshish> one node == fuel(internally multinode) + KB serivce
08:14:34 <SAshish> two things on same node
08:15:05 <SAshish> Am I making sense?
08:15:07 <sorantis> for two environments fuel will need two nodes
08:15:13 <joehuang123> If we have to setup the environment manually, then the FT can only be done manually
08:15:19 <sorantis> or we can use virtual fuel
08:15:35 <SAshish> that two nodes will be VMs created by fuel master/node
08:16:16 <sorantis> yes, but you’re refering to virtual box MOS
08:16:19 <SAshish> fuel creates multiple VMs for multiple environments
08:16:23 <SAshish> yes
08:16:25 <joehuang123> We tried two regions environment in vitual box
08:16:26 <SAshish> exactly
08:16:36 <SAshish> virtual box MOS
08:16:58 <joehuang123> what's MOS?
08:17:04 <sorantis> mirantis openstack
08:17:13 <SAshish> latest has come recently 8.0 liberty based
08:17:25 <joehuang123> got it
08:17:42 <sorantis> OPNFV has a different version i think
08:17:53 <SAshish> may be lower
08:18:25 <sorantis> so I guess first we need to check if we can use virtualbox MOS in opnfv for kb
08:18:39 <SAshish> one question here guys
08:18:58 <SAshish> this setup is for our FTs
08:19:21 <joehuang123> not installer project
08:19:37 <SAshish> if we push FTs when gerrit runs FTs?
08:19:56 <joehuang123> for installer, it'll be good to use virtual box MOS
08:20:09 <SAshish> it also needs some environment right?
08:20:39 <joehuang123> for FT, can we integrate into CI in gerrit
08:20:42 <SAshish> suppose I push my FTs for KB, do we need to tell gerrit where is the env details
08:21:03 <joehuang123> to Ashish, that's also my concerns
08:21:08 <SAshish> okay
08:22:03 <joehuang123> All current Function test is based on one openstack instance
08:22:28 <joehuang123> otherwise, we can just integrate it into function test projects
08:22:30 <SAshish> and from where they get that instance
08:22:45 <SAshish> that openstack instance?
08:23:05 <joehuang123> Not like CI always use the latest one
08:23:16 <joehuang123> OPNFV use the stable release
08:26:47 <SAshish> If we create installer then from where we get infrastructure
08:27:30 <SAshish> may be I have a very basic question.
08:28:37 <sorantis> ok
08:28:46 <sorantis> so I’ve just talked to opnfv CI guys here
08:28:48 <SAshish> see. our code is on gerrit, suppose we have a multisite installer, we need some infrastructure to install with our installer. this infrastructure has to be accessible by our code on gerrit
08:28:59 <sorantis> first we need to get a physical blade, I’ll try to do that this week
08:29:17 <SAshish> exactly.
08:29:26 <joehuang123> great sorantis
08:29:29 <sorantis> when we will have secured physical resources we can either use fuel to create two environments or use two fuels
08:29:42 <sorantis> they said both can be done
08:30:01 <sorantis> if we want to automate deployment of OpenStack, this can also be done, but later
08:30:06 <sorantis> so first thing first, get a blade
08:30:27 <sorantis> I will talk to Dan Smith and describe him our requirements.
08:30:48 <joehuang123> That's absolutely helpful
08:31:02 <sorantis> One important thing to keep in mind, is that that environment will not only be for KB, but for general multisite purpose
08:31:27 <SAshish> sure
08:31:36 <joehuang123> I think so, for example, multisite identity management
08:31:37 <sorantis> so in such case, perhaps having a static two region deployment is good enough for a start
08:31:49 <sorantis> yes
08:31:54 <joehuang123> Two regions at first is enough
08:32:01 <sorantis> we can actually automate the identity management use case too
08:32:38 <joehuang123> Support Fernet token?
08:33:04 <joehuang123> and clustered DB for the backend of KeyStone
08:33:26 <sorantis> yes
08:33:31 <sorantis> but first, the blade! :)
08:33:50 <SAshish> so, the access for the blade has to be there
08:33:53 <joehuang123> Yes, hope you can get one blade
08:34:13 <sorantis> again, that was just a suggestion from the CI folks, not a promise
08:34:15 <SAshish> so can it be accessed from gerrit/jenkins?
08:34:17 <sorantis> let me work on it
08:35:09 <SAshish> one more Q: for OPNFV do we need to get our code in review.opnfv.org
08:35:20 <SAshish> right now it is review.openstack.org
08:35:43 <joehuang123> now we leverage the infrastructure of openstack
08:35:50 <sorantis> yes
08:35:56 <sorantis> we don’t move the code anywhere
08:36:04 <SAshish> okay.
08:36:43 <joehuang123> I think OpenStack infrastructure is good for development
08:37:12 <SAshish> we need to request for infrastructure?
08:37:22 <joehuang123> I also talked to compass4nfv guys for the installation
08:37:35 <joehuang123> but currently, the most urgent one is FT
08:39:16 <SAshish> Yes, I want to verify the code till now once.
08:39:27 <SAshish> started looking into Tempest as well
08:39:29 <joehuang123> ok. It takes time for Dimitri to get the blade
08:39:32 <joehuang123> Ashish, how about the discussion with the locking?
08:39:37 <SAshish> sure
08:39:40 <SAshish> I have replied
08:40:06 <joehuang123> Dimitri, your opinion?
08:40:27 <SAshish> It's a little too heavy to use DB for locking. From my point of view, using a QuotaManager level variable to keep track the project list to be rebalanced, once it's rebalanced , remove it from the project list, if list become empty, then a new re balance cycle could be started when the timer comes. If the projectlist is not empty when the new timer comes, no action will be done, the previous timer is still there working until a
08:40:33 <SAshish> this was Joe's comment
08:40:47 <SAshish> I think its better to use DB for locking because of below reasons(2nd point is crucial).  1. Can be shared with any other sync job of Kingbird.  2. Align with other openstack projects in the way(using DB) locking  implemented there(Heat & Senlin). Current implementation in DB is based on locking in Senlin/Heat. I am sure other openstack projects also use DB for locking.  3. If DB, then at any instant of time, we can query if th
08:41:03 <SAshish> mine here
08:41:04 <joehuang123> #link https://review.openstack.org/#/c/292917
08:41:11 <SAshish> yeah better
08:41:14 <sorantis> I also would like to have locking
08:41:33 <sorantis> I think this mechanism will come in handing in many other situations
08:41:42 <sorantis> and it’s already there
08:41:44 <sorantis> so why not use it
08:41:49 <SAshish> which one?
08:41:52 <SAshish> locking with DB?
08:41:52 <sorantis> it’s actually very easy to use
08:41:54 <sorantis> yes
08:41:58 <sorantis> your commit
08:41:58 <SAshish> yes
08:42:01 <SAshish> fine
08:42:24 <joehuang123> ok
08:42:48 <joehuang123> but locking using under galera cluster has a little different behavior
08:43:35 <joehuang123> #link http://www.joinfu.com/2015/01/understanding-reservations-concurrency-locking-in-nova/
08:43:37 <SAshish> mysql behaviour ?
08:44:09 <sorantis> yes, I’ve read this one
08:44:13 <sorantis> good article
08:44:19 <joehuang123> mysql without and with galera is a little different
08:45:10 <joehuang123> have to catch "40001 dead lock error"
08:45:54 <joehuang123> no matter for primary key record duplicate writing or update a same record
08:46:15 <SAshish> in our case one sync task have access to lock table
08:46:50 <joehuang123> Ok. I'll review the patch next Monday, for tomorrow I'll leave office.
08:47:00 <sorantis> I’ve sent a request
08:47:08 <sorantis> will update you as soon as there’s news
08:47:20 <sorantis> is there anything else to discuss?
08:47:28 <SAshish> nothing.
08:47:29 <sorantis> I have to run quite soon
08:47:32 <joehuang123> no
08:47:38 <joehuang123> so let's end the meeting
08:47:45 <sorantis> joe please review when you have time
08:47:48 <joehuang123> thank you
08:47:53 <SAshish> yeah plese
08:47:53 <sorantis> thanks and bye
08:47:53 <joehuang123> #endmeeting