13:01:20 <joehuang> #startmeeting tricircle
13:01:20 <openstack> Meeting started Wed May  4 13:01:20 2016 UTC and is due to finish in 60 minutes.  The chair is joehuang. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:21 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:24 <openstack> The meeting name has been set to 'tricircle'
13:01:48 <joehuang> #topic rollcall
13:01:54 <joehuang> #info joehuang
13:02:09 <Ronghui_> #info Ronghui
13:02:19 <zhiyuan> #info zhiyuan
13:02:20 <Yipei> #info Yipei
13:02:37 <joehuang> #topic PTL election
13:03:40 <joehuang> hello, to be an openstack TC approved big tent project
13:03:49 <joehuang> we need to do PTL election
13:04:37 <joehuang> in OpenStack tradition, every one in the project could nominate himself to serve as the PTL of the release
13:04:59 <Ronghui_> what is the PTL?
13:05:29 <Yipei> project team leader?
13:05:36 <joehuang> project technical leader
13:06:01 <Ronghui_> that's cool
13:07:21 <joehuang> if you are interested in leading the project for the release, you can nominate yourself as the PTL, and members in the project can vote who will be the PTL if more than one nomination
13:07:22 <zhiyuan> I heard that there is a tool to run the election?
13:08:05 <joehuang> if more than one nomination happened, then we need to use the tool for election
13:09:22 <joehuang> so first a period for self nomination, then election
13:09:36 <Ronghui_> hi, all  we just update the spec for cross L2 networking
13:09:56 <Ronghui_> please review :)
13:10:11 <joehuang> will do, ronghui, thanks for the update
13:10:29 <joehuang> we can refer to https://wiki.openstack.org/wiki/PTL_Elections_March_2016 for the election procedure
13:11:09 <joehuang> we will also have one week for nomination, another week for election
13:12:37 <joehuang> how do you think that nomination period from May 9 ~ May 13, election from May 16 ~ May 20 if more than one nomination
13:12:54 <Ronghui_> no problem
13:13:02 <zhiyuan> that's fine
13:13:11 <Ronghui_> what we must to do?
13:13:15 <Yipei> fine for me
13:13:50 <joehuang> if you want to be the PTL for Newton release, than send self nomination letter in the mail-list
13:13:59 <Ronghui_> @zhiyuan  it can see the flavor list of Pod2 but the problem is also happen
13:14:03 <joehuang> will search one example
13:14:45 <zhiyuan> @Ronghui, ok, let's discuss this problem offline
13:14:56 <Ronghui_> ok
13:16:59 <joehuang> for example: http://permalink.gmane.org/gmane.comp.cloud.openstack.devel/78989
13:17:27 <joehuang> another self nomination: http://www.tesora.com/openstack-trove-ptl-candidacy/
13:18:26 <joehuang> #info PTL nomination period from May 9 ~ May 13, election from May 16 ~ May 20 if more than one nomination
13:18:52 <joehuang> #info if you want to be the PTL for Newton release, then send self nomination letter in the mail-list
13:19:35 <joehuang> ok, will send a mail tomorrow in m-l for the PTL election
13:19:57 <joehuang> next topic server group
13:20:06 <joehuang> #topic server group
13:20:41 <joehuang> hi, zhiyuan, in the virtual design summit, the server group topic is not finished yet, let's discuss about it
13:20:47 <joehuang> please
13:21:13 <zhiyuan> ok, two problems for server group
13:22:01 <zhiyuan> one is that if two VMs are created in two pods, we cannot guarantee server group affinity
13:22:53 <joehuang> for server group with affinity policy, should be forwarded to same pod in Tricircle
13:22:56 <zhiyuan> the other one is when to create server group in bottom pod. create server group when creating VM, or when create server group in top pod
13:23:35 <joehuang> I saw that for server group only CRD, no U operation
13:23:58 <joehuang> that means the once the policy is defined, no change will be done on the same group
13:24:53 <joehuang> and currently only two policy supported for server group: affinity, anti-affinity
13:24:59 <zhiyuan> so if the pod binding changed, we reject creating VM in the server group which already has VMs, right?
13:25:32 <joehuang> do you mean active pod binding?
13:26:13 <joehuang> currently the assumption that only one pod binding in one AZ needs to be updated if we support multiple pod in one AZ
13:26:51 <joehuang> only one "active" pod binding, but can multiple inactive pod binding in the history
13:27:25 <zhiyuan> yes, but the previous VM in the same server group may be created in the previous pod
13:27:49 <joehuang> so do we need server group pod binding?
13:28:08 <joehuang> that means a server group will only be created in one pod
13:28:42 <joehuang> or do we need to support server group across multiple pods for anti-affinity policy?
13:29:09 <joehuang> for affinity policy, all VMs should allocated in same pod
13:29:58 <zhiyuan> i think we can support anti-affinity policy.
13:30:12 <joehuang> across multiple pods?
13:30:14 <Yipei> does it mean we need a mapping between server group and pod?
13:30:20 <Ronghui_> u mean the pod instance of the tradition compute node in the server group?
13:30:30 <zhiyuan> for affinity policy, we need to check if the new VM is in the same pod with the old VMs in the same server group
13:30:55 <joehuang> pod is one bottom openstack instance
13:31:10 <joehuang> agree for affinity policy
13:31:30 <zhiyuan> if AZ-pod binding changed, the new VM cannot be placed in the same pod with the old one, thus creation fails
13:32:37 <zhiyuan> ok, then what about the creation time for server group?
13:32:41 <joehuang> that usually means no more capacity to create VM in the old pod
13:32:56 <Ronghui_> and  there will be two layer server group
13:33:14 <joehuang> to Ronghui, yes
13:33:20 <Ronghui_> one for pod and one for compute node in same pod
13:33:27 <zhiyuan> "that usually means no more capacity to create VM in the old pod" +1
13:34:37 <joehuang> not "one for pod"
13:35:32 <Ronghui_> "that usually means no more capacity to create VM in the old pod" that means  we not only can not create VM in the same node but also can not create the VM in the same bottom openstack?
13:36:10 <joehuang> to Ronghui, what do you mean "same node"
13:36:54 <Ronghui_> same compute node in bottom openstack  the tradition server group
13:37:11 <joehuang> to affinity policy, yes
13:37:54 <joehuang> ok, zhiyuan, now we have some consensus on server group, please have a summary here
13:38:18 <Ronghui_> wheather the commutication cost is too big for this case between affinity VMs
13:40:12 <joehuang> usually need better communication capability or a close group relationship of VMs
13:40:48 <Ronghui_> agree
13:40:58 <zhiyuan> for anti-affinity policy, VMs in the same server group needs to be placed in the same bottom pod, if AZ-pod binding is changed, creating a new VM in a server group that already has VMs will fail
13:41:08 <zhiyuan> sorry for affinity policy
13:41:46 <joehuang> #info  for affinity policy, VMs in the same server group needs to be placed in the same bottom pod, if AZ-pod binding is changed, creating a new VM in a server group that already has VMs will fail
13:43:01 <zhiyuan> for anti-affinity policy, AZ-pod binding change has not impact, since creating VMs in different bottom pods also satisfies anti-affinity policy
13:43:12 <joehuang> yes
13:43:27 <joehuang> #info anti-affinity policy, AZ-pod binding change has not impact, since creating VMs in different bottom pods also satisfies anti-affinity policy
13:44:22 <zhiyuan> ok, what about the time to create server group?
13:44:23 <joehuang> ok next topic
13:44:30 <joehuang> sorry
13:44:48 <joehuang> if no routing info for a server group
13:45:19 <joehuang> when first time to create a VM in the server group
13:46:48 <joehuang> when create a VM in the server  group and find no regarding routing info in the binding pod
13:49:25 <joehuang> ok, please vote Shinobu to be a core reviewer or not in the mail-list
13:49:57 <joehuang> and yipei, how about dynamic pod binding?
13:51:19 <Yipei> For the code, I still study how to write unit test. For the pod binding framework, I plan to submit a patch before next weekend.
13:51:33 <Yipei> i think i miss something about server group
13:51:37 <joehuang> quite good
13:51:40 <joehuang> please
13:51:58 <Yipei> For the spec file, I have already add some content, but still need to organise and refine the draft. I will update it before this weekend.
13:52:26 <joehuang> your question about server group?
13:53:51 <Yipei> no for now, i think i did not consider the server group issues
13:54:43 <joehuang> ok, refer to http://developer.openstack.org/api-ref-compute-v2.1.html#os-server-groups-v2.1 for server group
13:55:15 <joehuang> ronghui just said that L2 networking spec updated, so let's review the spec
13:55:22 <Yipei> ok, got it
13:55:27 <zhiyuan> ok
13:55:45 <joehuang> ok, thank you for attending the meeting
13:55:52 <joehuang> it's time to end the meeting
13:56:06 <joehuang> thank you
13:56:12 <joehuang> #endmeeting