06:01:30 <yoctozepto> #startmeeting masakari
06:01:30 <opendevmeet> Meeting started Tue Jun 15 06:01:30 2021 UTC and is due to finish in 60 minutes.  The chair is yoctozepto. Information about MeetBot at http://wiki.debian.org/MeetBot.
06:01:30 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
06:01:30 <opendevmeet> The meeting name has been set to 'masakari'
06:01:36 <yoctozepto> #topic Roll-call
06:01:38 <yoctozepto> \o/
06:02:01 <suzhengwei__> o
06:02:12 <XinxinShen> o
06:02:29 <jopdorp> o/
06:03:24 * yoctozepto glad to have so many participants
06:03:48 <yoctozepto> #topic Agenda
06:04:04 <yoctozepto> * Announcements
06:04:04 <yoctozepto> * Review action items from the last meeting
06:04:04 <yoctozepto> * CI status
06:04:04 <yoctozepto> * Backports pending reviews
06:04:04 <yoctozepto> * Xena planning -> https://etherpad.opendev.org/p/masakari-xena-ptg
06:04:05 <yoctozepto> * Open discussion
06:04:12 <yoctozepto> #topic Announcements
06:04:15 <yoctozepto> none
06:04:34 <yoctozepto> I have to change the approach to agenda so that I don't include these mostly unused topics
06:04:50 <yoctozepto> #topic Review action items from the last meeting
06:04:52 <yoctozepto> none
06:04:55 <yoctozepto> (again)
06:05:04 <yoctozepto> #topic CI status
06:05:24 <yoctozepto> rolling green
06:05:36 <yoctozepto> #topic Backports pending reviews
06:05:45 <yoctozepto> none
06:06:00 <yoctozepto> #topic Xena planning -> https://etherpad.opendev.org/p/masakari-xena-ptg
06:06:20 <yoctozepto> all right
06:06:27 <yoctozepto> the main point of our meetings
06:06:29 <yoctozepto> Xena
06:06:36 <yoctozepto> let's see the schedule
06:06:47 <yoctozepto> #link https://releases.openstack.org/xena/schedule.html
06:07:00 <yoctozepto> it's R-16
06:07:43 <yoctozepto> next week storage projects freeze their specs, we should definitely decide on what we implement this cycle as well
06:08:02 <jopdorp> Ok
06:08:09 <yoctozepto> I think we need to focus on adding support for consul
06:08:14 <yoctozepto> what do you think?
06:08:26 <jopdorp> Sounds like a plan
06:08:52 <jopdorp> suzhengwei__: what do you think?
06:09:02 <yoctozepto> to be honest, I was mostly asking suzhengwei__ ;-)
06:09:30 <suzhengwei__> well, I will update it next two week.
06:09:59 <yoctozepto> ok, wonderful
06:10:08 <yoctozepto> I will review what we have so far to better understand it
06:10:42 <yoctozepto> #link
06:10:45 <yoctozepto> #link https://etherpad.opendev.org/p/masakari-xena-ptg
06:11:21 <yoctozepto> "host monitor by consul"
06:14:32 <yoctozepto> ok, what else?
06:14:44 <yoctozepto> (regarging new Xena features)
06:14:48 <yoctozepto> regarding*
06:15:24 <suzhengwei__> "reduce failover time", there are some new comments.
06:17:36 <yoctozepto> suzhengwei__: thanks, commented
06:23:25 <yoctozepto> good point about scheduling
06:23:36 <yoctozepto> I added even more thougths now
06:31:31 <suzhengwei> I will reconsider the comments and update the patch.
06:32:11 <yoctozepto> thanks
06:32:24 <yoctozepto> anything else on Xena?
06:33:55 <jopdorp> I think I will add a spec for migrating instances back to the hosts from which they were evacuated
06:34:36 <yoctozepto> good
06:35:52 <suzhengwei> A question, when openstack on k8s, the hostname in nova-compute service(pod) has a random suffix.
06:36:26 <yoctozepto> can this be disabled?
06:36:40 <suzhengwei> Every time the pod is rescheduled or restart, then new hostname.
06:36:58 <yoctozepto> oh, bad
06:37:07 <jopdorp> But part of the name is always the same?
06:37:18 <suzhengwei> yes.
06:37:30 <jopdorp> Should we add regex matching to support this type of deployment?
06:37:43 <yoctozepto> we can't regex into nova
06:37:58 <jopdorp> We can list the hosts
06:38:11 <yoctozepto> performance not happy :-)
06:38:14 <jopdorp> Then filter the matches
06:38:24 <jopdorp> Well
06:38:27 <yoctozepto> is that the behaviour of openstack-helm?
06:38:31 <jopdorp> Su has a couple of hundred
06:38:52 <yoctozepto> the assumption of node names is that they provide some provenance on host usage
06:39:00 <yoctozepto> so it might break multiple systems anyhow
06:39:14 <yoctozepto> if they are randomised
06:39:34 <yoctozepto> the host name part is configurable
06:39:53 <yoctozepto> openstack helm must be ensuring already that there are no two nova computes on one kubelet
06:40:05 <yoctozepto> so it could take the hostname from kubelet
06:40:14 <yoctozepto> which does not change
06:40:17 <yoctozepto> problem solved
06:40:24 <yoctozepto> (this and many others)
06:41:07 <jopdorp> You want to adjust the deployment system that su uses, or do you want to Integrate masakari with it?
06:42:01 <jopdorp> As I understand, you want to adjust it to stabilize the host name
06:42:03 <yoctozepto> I want it to be fixed to know the assumptions
06:42:24 <yoctozepto> any other deployment system is aware
06:43:03 <yoctozepto> telemetry and monasca will be unhappy about random names as well
06:43:13 <yoctozepto> as it breaks the node history
06:43:29 <jopdorp> I wonder why it uses a randomized suffix now
06:43:48 <yoctozepto> because that's what you get in k8s pods as hostname I guess
06:44:14 <jopdorp> Because of default k8s pod id generation behavior
06:46:38 <yoctozepto> #topic Open discussion
06:46:50 <jopdorp> We'd have to look for all places where the host name is used, and map it to the correct pod with a lookup of the id
06:47:33 <jopdorp> Better discussed in openstack helm development channel
06:47:36 <yoctozepto> just make openstack helm configure stable names, it's easier on its side
06:47:38 <yoctozepto> indeed
06:48:19 <suzhengwei> yep. I think there is solution to make the hostname stable and I need to find it.
06:48:53 <suzhengwei> thanks
06:49:17 <yoctozepto> sure thing, you are always welcome
06:51:35 <jopdorp> I don't have anything else to discuss
06:51:54 <yoctozepto> me neither
06:52:15 <yoctozepto> suzhengwei: anything else for the last 8 minutes?
06:52:20 <suzhengwei> me neither
06:52:28 <yoctozepto> ok
06:52:36 <yoctozepto> thank you all for attending today
06:52:39 <yoctozepto> #endmeeting