08:04:42 <joehuang> #startmeeting multisite
08:04:42 <collabot> Meeting started Thu Sep 10 08:04:42 2015 UTC.  The chair is joehuang. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:04:42 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:04:42 <collabot> The meeting name has been set to 'multisite'
08:04:45 <uli-k> Tapio_T: huomenta!
08:05:10 <zhipeng> Chinese is getting out of control lol
08:05:39 <joehuang> #topic rollcall
08:05:49 <joehuang> #info joehuang
08:05:53 <colintd> #info colintd
08:06:03 <Xiaolong> #info Xiaolong
08:06:03 <fuqiao> #info fuqiao
08:06:04 <sorantis> #info sorantis
08:06:08 <joehuang> uli, you are language expert.Even Finish?
08:06:10 <zhipeng> #info zhipeng
08:06:21 <Tapio_T> #info Tapio_T
08:06:32 <uli-k> #info Uli Kleber
08:06:44 <joehuang> #info agenda
08:06:49 <MatthewLi> #info Jun Li
08:06:57 <joehuang> #info we have 3 agenda today
08:07:10 <zhipeng> Octopus team is crushing the party
08:07:14 <joehuang> #info Jira issue follow up
08:07:36 <joehuang> #info 2 collboration with HA team
08:07:47 <joehuang> #info 3 use case 4 discussion
08:07:59 <joehuang> let's start from the topic 1
08:08:06 <sorantis> ok
08:08:22 <joehuang> #info we have 8 issues registered now
08:08:37 <joehuang> and most of them had one assignee
08:08:44 <sorantis> so, I’ve added some example characteristics and requirements to the etherpad
08:09:06 <joehuang> yes, I think we can close these kind of issue ticket
08:09:23 <sorantis> I’ll just set it to complete
08:09:24 <joehuang> and just leave the gerrit review ticket there
08:09:33 <joehuang> yes, please
08:10:09 <joehuang> and according to the project manager Debra's requirement
08:10:34 <joehuang> our output document jira issue should be marked with critical and link to B label or epic
08:10:45 <sorantis> #info MULTISITE-5  resolved
08:12:26 <joehuang> and let's focus on the review and approve for the output doc bp bug
08:12:26 <joehuang> So after you finish your short term issue, please close it
08:12:37 <joehuang> #info I just updated the commiter list, Dimitri ( sorantis ) is added to the list of committer
08:13:10 <joehuang> and it's approve by aric and merged yesterday
08:13:29 <joehuang> any question on the topic 1 , jira issue
08:13:32 <sorantis> great, thanks joe
08:13:41 <sorantis> any other updates on issues?
08:13:55 <joehuang> on jira and issues
08:13:55 <sorantis> what about solution proposal 2 and 3?
08:13:59 <zhipeng> colin will work on use case 2 right ?
08:14:30 <joehuang> Zhipeng upload use case 3 .rst for review in gerrit
08:14:54 <colintd> Yes, just having some tool problems at present which I'm sorting with Joe.  Aim to close this week.
08:14:59 <joehuang> for use case 2, we need some more work on .rst, I will work together with colin
08:15:38 <joehuang> and it would be great for people to co-work on same doc together
08:16:06 <sorantis> sure, which doc exactly are you refering to?
08:16:31 <joehuang> any one, volunteering encouraged
08:17:22 <joehuang> and Bin Hu asked me to present the BP/bug we are in  working
08:17:35 <joehuang> I'll try to present today in His meeting
08:19:16 <joehuang> ok, let's talk about the topic 2
08:19:25 <joehuang> currently the second one need more work
08:19:29 <joehuang> #topic colloboration with HA team
08:19:39 <joehuang> Hi, Qiao
08:19:42 <fuqiao> hi
08:19:56 <joehuang> as what we discussed in the mail-list
08:20:09 <joehuang> we want to bridge these two teams
08:20:15 <joehuang> your ideas
08:20:29 <fuqiao> the HA team is actually working on some HA usecase and scenario analysis doc
08:20:49 <fuqiao> what we are thinking is, we also would like to address the work multisite is doing
08:21:03 <fuqiao> and refer it as multisite usecases
08:21:13 <joehuang> but not identify the use case work in one site or multi-sites?
08:21:17 <fuqiao> so we would like to know about your progress
08:21:39 <fuqiao> I can give you the link.
08:21:54 <fuqiao> https://gerrit.opnfv.org/gerrit/#/c/1072/
08:22:01 <fuqiao> for now, we only work on basic usecases
08:22:27 <fuqiao> we will have another section, including some usecases that may be more complicated, like multisite
08:23:20 <fuqiao> we want to avoid repeat work here, so we think if you have done the usecase analysis work, we would be happy to join the discussion, give some comments, and also refer to your work in our doc.
08:23:24 <joehuang> for multisite, we discussed use case: 1.multisite identity service management, 2. VNF high availability across VIM, 3.VNF Geo-site Redundancy
08:24:01 <fuqiao> for example, we can have a new section about multisite, but we will refer to your project's work about this scenario.
08:24:21 <fuqiao> yes, I have read through the slides you shared. there are some questions from our team.
08:24:34 <fuqiao> I haven't seen the answers from you team:)
08:24:34 <joehuang> good
08:25:06 <joehuang> which one. Could you comment on the gerrit review, it would be addressed one by one
08:25:12 <fuqiao> one thing I think I want to know now is how is the progress of these usecase work
08:25:31 <fuqiao> yes, sure, I can put the comments on gerrit. It will be much easier for the discussion
08:25:46 <fuqiao> are you planning to have the usecase analysis work to be output with release B?
08:25:55 <zhipeng> the 3 you mentioned are all concluded in Multisite project discussion
08:26:18 <zhipeng> therefore we are generating requirements from these agreed use cases
08:26:19 <fuqiao> OK  great. Then I will check on the gerrit and see if we have further questions.
08:26:26 <joehuang> we would like it be in release B
08:26:43 <fuqiao> OK. our usecase doc is also planning to join release B
08:26:53 <fuqiao> All the 3 usecases are on gerrit right?
08:27:04 <zhipeng> 1 and 3 are on gerrit
08:27:11 <zhipeng> 2 is wip, but soon
08:27:16 <fuqiao> OK
08:27:20 <zhipeng> should be done before end of this week
08:27:36 <fuqiao> Great. We will provide comments and feedbacks on the gerrit
08:27:58 <fuqiao> I think from the slides, you are discribing the usecase, then provide requirements on openstack.
08:28:08 <joehuang> yes
08:28:24 <fuqiao> I think, for the HA team, we may also provide requirement on other components apart from openstack
08:28:42 <fuqiao> do you want to cover other components or only openstack?
08:29:05 <joehuang> #link https://wiki.opnfv.org/_media/multisite/multisite_bp-bug_review_0910.pdf
08:29:15 <joehuang> only openstack
08:29:36 <joehuang> other components is considered to be out of scope of the project
08:29:45 <fuqiao> OK. Then I think we can also talk about requirement of other components in the HA usecase doc, if there is any
08:29:50 <fuqiao> OK got it
08:30:18 <fuqiao> Then I think I can put some action bullet for the HA team
08:30:30 <joehuang> good
08:30:35 <fuqiao> #info action 1: review multisite gerrit and provide comments
08:31:05 <fuqiao> #info action 2: analyse multisite usecases, include it in the usecase doc, and refer to the multisite project work
08:31:07 <joehuang> hi team, what we can do for HA team
08:31:09 <zhipeng> joehuang do we also need to reference HA team's doc ?
08:31:51 <fuqiao> #info action 3: analyse multisite usecases, see if we need to add any other requirements for the other components other than openstack
08:31:58 <joehuang> do you have seperate doc for each use case, or collected in one
08:32:05 <fuqiao> collect in one
08:32:30 <fuqiao> but you can refer to the doc and for each usecase, we have a seperate section
08:32:38 <fuqiao> so you can refer to a specific section
08:32:58 <joehuang> bookmark for each section
08:33:28 <zhipeng> maybe also on JIRA to have dependencies of two projects
08:34:00 <fuqiao> please also visit the HA project gerrit and see if you have any comments on the usecase doc
08:34:13 <fuqiao> we will forward the multisite related part to you once finished
08:34:38 <fuqiao> we haven't buit a jira for the multisite related doc. Do you already have one?
08:34:45 <joehuang> good
08:35:20 <joehuang> no doc in jira, but in gerrit
08:35:48 <fuqiao> another thing is that, in your usecase 3, you said about snapshot. is it a common solution for multisite VNF HA?
08:35:57 <zhipeng> usecase 3 is for geo redundancy
08:36:01 <joehuang> it's used to save, but not live active/standby or active/active
08:36:05 <fuqiao> because we think this might have some extra requirement on VNF. I don't know if currently their is any actually practice for such multisite deployment of VNF
08:36:16 <joehuang> it's used to save VNF, but not live active/standby or active/active
08:36:32 <zhipeng> fuqiao this is the difference between multisite and HA I think
08:36:33 <fuqiao> but you have to froze the VNF in order to save, right
08:36:50 <zhipeng> where we only consider standby-standby scenarios for now
08:36:57 <joehuang> the option 3 will not freeze
08:37:06 <joehuang> option 1/2 will
08:37:12 <fuqiao> what do you mean by standby-standby
08:37:41 <fuqiao> OK. It seems I really need to read your gerrit asap. I actually don't know your option 1/2/3...
08:38:11 <fuqiao> it's OK. I will read the gerrit and see if I can understand
08:38:21 <sorantis> https://etherpad.opnfv.org/p/VNF_Geo_site_redundancy
08:38:28 <zhipeng> active standby..
08:38:42 <zhipeng> sorry was messed around with lot of stuff
08:39:16 <fuqiao> still confused, we also consider active standby. what is the difference?
08:39:24 <colintd> On the whether that is practice for multisite deployments, we're already working with carriers who are using both VNF with active:standby elements splits between 2 local clouds, and active:active GR between remote clouds.  All of this uses application awareness, as opposed to infrastructure implemented function
08:40:31 <fuqiao> when you say, applicatoin awarness, do you mean the application should know of the snapshot thing?
08:40:45 <fuqiao> so it knows it will be frozen from time to time?
08:40:49 <joehuang> The application should know the replication
08:40:52 <colintd> For those customers, HA is intimately tied up with multisite.  Call maintaining (classic HA) for close clouds where we can do media failover, disaster recovery / geographic redundancy for less connected clouds
08:41:06 <joehuang> for option 3, no freeze will happen
08:41:27 <colintd> We don't use snapshots, it is application level replication couples with backend replication of specific datastores by the infrastructure
08:42:01 <joehuang> Colin is refer to multi-master cluster running over multiple cloud and work at the same time
08:42:14 <fuqiao> OK
08:42:26 <fuqiao> This makes sense to me
08:42:58 <fuqiao> So snapshot is not a necessary action right?
08:43:30 <joehuang> Geo_site_redundancy has two way of meaning, one is both working( dual-live), and another is cold backup
08:43:46 <fuqiao> in your practise, you are relying on the application level replication, as we also have for VNFs deployed in the same site
08:44:18 <joehuang> yes some scenario is like this
08:45:25 <fuqiao> OK. So if I understand right, you have extra requirement of openstack, if the VNF is deployed on two sites rather than 1, and there will be application level replication.
08:45:56 <joehuang> yes
08:46:14 <fuqiao> OK. I will read though your doc to understand more:0
08:46:29 <joehuang> the last one is the better one, for these application is cloud ready
08:46:58 <fuqiao> got it
08:47:08 <joehuang> but for some applications which are not cloud-ready. snapshot will be used
08:47:20 <fuqiao> OK.
08:47:33 <fuqiao> but do you consider the outage of the app for the snapshot?
08:47:53 <fuqiao> this is what we are concerned of when talking about the snapshot
08:48:52 <joehuang> 's responsibility
08:49:06 <fuqiao> ?
08:49:34 <joehuang> for how to achieve higher availability, it will be application level/and upper level of OpenStack
08:49:34 <fuqiao> don't understand. whose responsibility?
08:49:42 <joehuang> we are looking at what's missing in OpenStack
08:50:13 <joehuang> for example, to collect the fault or emergency alarm and reaction to switch over or autoscaling
08:51:26 <fuqiao> you suppose snapshot is done by openstack or by upperlayer?
08:52:00 <joehuang> the trigger is upper layer
08:52:12 <fuqiao> OK. Then I got it
08:52:46 <joehuang> so let's have a few minutes on topic 3?
08:52:53 <sorantis> NFVO?
08:53:01 <fuqiao> you are supposing, the upper layer should do either snapshot, or application replication, so as to provide HA for the app. While your project only consider what kind of capability openstack need to support upper layer
08:53:02 <joehuang> #topic use case 4
08:53:16 <fuqiao> OK please goto next topic
08:53:23 <fuqiao> sorry for taking so much time
08:53:26 <joehuang> to fuqiao, yes
08:53:47 <fuqiao> I will have our team read through your doc and see if we have further comments
08:53:49 <fuqiao> Thank you
08:53:49 <joehuang> will NFVO in charge all of these?
08:54:00 <joehuang> thanks qiao
08:54:15 <sorantis> that was my question. This can be NVFO, or any OSS
08:54:19 <fuqiao> I don't know, in the HA project, we assume their should be an HA manager responsible for the HA
08:55:09 <joehuang> use case 4: Use case 4. Centralized service for resources management and/or replication (sync tenant resources like images, ssh-keys, security groups, etc)
08:55:10 <fuqiao> NFVO might be too high level for governing the VNF
08:55:50 <joehuang> to qiao, you mean the management of HA?
08:55:50 <sorantis> so for use case 4 I’ve started drafting a solution proposal for centralized quota management
08:56:07 <sorantis> https://etherpad.opnfv.org/p/centralized_quota_management
08:56:15 <joehuang> great
08:56:17 <fuqiao> yes
08:56:22 <sorantis> I’ve done some work on it already
08:56:41 <sorantis> basically the idea is to bridge the gap for quota management in OpenStack(s)
08:56:58 <sorantis> I’ll add more information to the doc and we can discuss it next time
08:57:12 <sorantis> mind you, for now it’s only quota related
08:57:22 <zhipeng> nice
08:57:28 <joehuang> how about other parts
08:57:42 <joehuang> would you plan to add them too
08:58:06 <sorantis> the other parts I’d rather first discuss with you
08:58:43 <joehuang> may we discuss in M-L first, time is almost over for the meeting
08:58:47 <sorantis> the difference between quota sync and, say, image sync is that quotas must be synced
08:58:53 <sorantis> images - probably
08:59:09 <sorantis> sure
08:59:20 <sorantis> have a look at the doc
08:59:24 <joehuang> sure
08:59:36 <sorantis> and we can continue on ml if we want to merge the other parts into this
08:59:37 <joehuang> let's discuss next meeting
08:59:49 <joehuang> and m-l
09:00:06 <joehuang> thanks for all of you attending the meeting
09:00:12 <joehuang> see you next time
09:00:23 <joehuang> #endmeeting