08:17:12 <joehuang> #startmeeting multisite
08:17:12 <collabot`> Meeting started Thu Feb 25 08:17:12 2016 UTC.  The chair is joehuang. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:17:12 <collabot`> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:17:12 <collabot`> The meeting name has been set to 'multisite'
08:17:13 <SAshish> Hi DImitri
08:17:34 <joehuang> #topic rollcall
08:17:38 <joehuang> #info joehuang
08:17:43 <SAshish> #info Ashish
08:17:50 <zhiyuan> #info zhiyuan
08:17:59 <zhipeng> #info zhipeng
08:18:36 <sorantis> #info dimitri
08:18:39 <sorantis> hi
08:18:42 <joehuang> Hello, just one information to share
08:19:22 <joehuang> before the meeting, just talked that the Glane community admit the use case for image replication between regions are legitimate use case
08:19:50 <Malla> #info Malla
08:20:03 <joehuang> Hopefully this could be started in N release of OpenStack
08:20:10 <sorantis> in Glance?
08:20:38 <joehuang> Glance will provide "image cloning", "image import" and "image export"
08:21:01 <sorantis> ok, so in Kingbird
08:21:07 <joehuang> Now the "image import" is in refector
08:21:30 <joehuang> Kingbird can mainly focus on controll plane, but not to redirect the image bits
08:21:56 <sorantis> what do you mean
08:22:03 <joehuang> For example, replicate one image to n Regions, then clone this image to n Glance
08:22:24 <sorantis> yes
08:22:32 <joehuang> the image data bits will directly transfer from one Glance to others
08:22:45 <sorantis> export-import
08:22:51 <joehuang> otherwise, Kingbird has to touch images bits, it's awful
08:22:51 <SAshish> that will be provided by glance you are telling rigt?
08:22:58 <joehuang> Yes
08:23:12 <sorantis> fair enough
08:23:14 <SAshish> replicate image to n regions
08:23:18 <joehuang> But need developer to work on it, use case is accepted by one core
08:23:44 <SAshish> then it will not be part of Kingbird
08:24:08 <joehuang> I think it's good news at least, I'll forward the mail discussion to you
08:24:26 <SAshish> sure
08:24:33 <joehuang> There are still some work in Kingbird, but more lightweight
08:25:29 <joehuang> and will not become the bottleneck of data plane for the image data transfer
08:25:40 <joehuang> in Kingbird
08:26:04 <sorantis> yes
08:26:18 <joehuang> #topic environment setup in release C
08:26:19 <sorantis> but first let’s see if Glance actually implements it
08:26:25 <joehuang> yes
08:26:57 <zhipeng> we need to define a limit set of features of Kingbird we want to bring in Rel C
08:27:04 <joehuang> How do you guys think about to setup a multisite environment in Release C. Currently all silo OpenStack installation
08:27:17 <zhipeng> from past experience more conservative the better
08:27:34 <joehuang> To zhipeng, In release C quota management focus first
08:27:44 <sorantis> +1
08:27:50 <SAshish> yes
08:27:56 <sorantis> I think we can aim QM for Mitaka in OpenStack
08:28:04 <sorantis> and then for C
08:28:08 <joehuang> yes
08:28:15 <zhipeng> agree
08:28:16 <joehuang> maybe preview release
08:28:30 <zhipeng> however we do need to consider the test env for C
08:28:51 <zhipeng> we need installers to support Multisite installation, and testing projects to support the testing
08:29:03 <sorantis> but before introducing Kingbird as a component in OPNFV we need to make sure there’s a MR environment in OPNFV
08:29:05 <zhipeng> and we need to do it early on
08:29:05 <joehuang> maybe setup an test env ready for later release
08:29:24 <joehuang> "MR"?
08:29:29 <sorantis> multi-region
08:29:34 <joehuang> ok
08:30:07 <zhipeng> yes need to talk to test team about that
08:30:08 <joehuang> yes MR env ready before introducing KB
08:30:53 <sorantis> more than kingbird we need to see if our other use cases are working
08:31:10 <joehuang> How many test labs/pods available now
08:31:14 <SAshish> multisite usecases
08:31:24 <sorantis> e.g. we have a design proposal for setting up centralized keystone
08:31:32 <joehuang> yes, at least identity management
08:31:49 <sorantis> so as a first step we can say:
08:31:52 <zhipeng> for test labs, i think 3-4 labs are working
08:31:52 <joehuang> distribued KeyStone
08:32:02 <sorantis> - set up mutisite env
08:32:04 <zhipeng> numbers of pods varies per lab
08:32:15 <sorantis> - set up centralized identity management
08:33:05 <joehuang> to Dimitri, which token format for KeyStone
08:33:46 <sorantis> our recommendation was to use PKI, right?
08:33:53 <joehuang> Fernet
08:33:53 <zhipeng> fernet?
08:34:07 <sorantis> then Fernet
08:34:12 <joehuang> Fernet, for not so big like PKI
08:34:32 <joehuang> PKI could be used in limited regions
08:35:25 <zhipeng> one question related to the lab setup
08:35:34 <zhipeng> do we need to have a real multisite env?
08:35:44 <zhipeng> for example to use Pharos projects
08:36:01 <zhipeng> to have OPNFV labs across the world deploy Multisite
08:36:06 <joehuang> Does Pharos establish MR env?
08:36:21 <zhipeng> no at the moment it just connects the labs i think
08:36:40 <zhipeng> Multisite would be first to actually do so, if we decide to go this way
08:37:01 <zhipeng> although I'm not favor of this, since this would be too much of work
08:37:06 <zhipeng> coordination and such
08:37:28 <joehuang> yes. Would test team to verify this?
08:38:02 <zhipeng> i will ask around Huawei OPNFV test teams to see what support we could get
08:38:28 <zhipeng> I would recommend to do it in one lab, and setup a MR env
08:38:55 <sorantis> I’m ok with that
08:39:23 <sorantis> we don’t necessarily need multiple labs, we don’t test any WAN conncectivity related use cases
08:39:38 <joehuang> agree, in one lab is acceptable
08:39:56 <zhipeng> sorantis yes
08:40:14 <joehuang> so Zhipeng, pls contact test team first to see if it's possible
08:40:29 <zhipeng> joehuang sure
08:40:59 <joehuang> #action contact test team for multi-region env. establishment and test
08:41:51 <joehuang> so let's discuss this later after response we get
08:42:58 <joehuang> One more topic is about the volume replication
08:43:20 <joehuang> The Cinder team has re-design the volume replication
08:43:48 <joehuang> #link https://specs.openstack.org/openstack/cinder-specs/specs/mitaka/cheesecake.html
08:44:08 <joehuang> #link https://review.openstack.org/#/c/275797/
08:46:09 <sorantis> this needs some proper reading :)
08:46:11 <joehuang> From the new design, we can renew the thinking in the 3rd option for VNF geo-site disaster recovery
08:47:05 <joehuang> yes, it's take time
08:47:16 <joehuang> it takes time to read
08:49:40 <SAshish> preservation of user data
08:50:09 <SAshish> so, what about the actual data?
08:50:37 <joehuang> the actual data is replicated by the backend
08:52:12 <joehuang> the patch was just uploaded Feb.4, and still in review process,
08:52:36 <sorantis> if merged, will it appear in M release?
08:53:22 <joehuang> I am not sure it can be merged, it's very limited time for M release, and it's only uploaded recently
08:53:24 <SAshish> yes. I guess
08:53:31 <SAshish> https://blueprints.launchpad.net/cinder/+spec/replication
08:53:39 <joehuang> It's a quite big refector
08:53:48 <SAshish> not sure..
08:53:48 <joehuang> refactor
08:55:10 <joehuang> Please comment in this patch if needed
08:55:40 <joehuang> This patch was submitted by previous cinder PTL
08:56:51 <joehuang> So let's see if this patch can fulfill Multisite GR requirement
08:58:26 <joehuang> And last one, currently configuration sample file in Kingbird is maintained mannually
08:59:04 <joehuang> We can use oslo tools to generate the configuration sample file, but need some code and a patch
08:59:31 <sorantis> yes
08:59:50 <sorantis> I’ll have a look
09:00:00 <joehuang> We can do it later
09:00:20 <SAshish> okay.
09:00:28 <joehuang> just let's add it to the todo list to reminder
09:00:38 <joehuang> ok
09:00:43 <joehuang> other topics?
09:00:50 <SAshish> no from my side.
09:01:05 <sorantis> nope
09:01:08 <SAshish> Plese reviewt the commits
09:01:14 <joehuang> will do
09:01:28 <joehuang> so let's conclude the meeting
09:01:40 <joehuang> #endmeeting