Jay, On point number 2 I proposed https://blueprints.launchpad.net/nova/+spec/vmware-auto-inventory soon after we merged the Cluster patch because of just the issues you are highlighting. I may have named it badly and it's purpose maybe outdated now... so I should definitely work on clarifying things like that.
My original plan for what I called "auto-inventory" was to enable the VCDriver to pull out individual ESX hosts and present them as nodes to Nova. So I mention this just to say, many of us have the same ideas but we've not acted on them yet. I think the time will be right to start working on some of these issues in Juno. Talk to you Wednesday. We won't finish the discussion then, but we can at least start having them. On Sat, Mar 22, 2014 at 8:22 AM, Jay Lau <jay.lau....@gmail.com> wrote: > Thanks Shawn, I have updated the title with VMWare. > > Yes, I know that live migration "works". But the problem is when a cluster > admin want to live migrate a VM instance, s/he will not know the target host > where to migrate, as s/he cannot get target host from nova compute because > currently VCDriver can only report cluster or resource pool as hypervisor > host but not ESX server. > > IMHO, the VCDriver should support live migration between cluster, resource > pool and ESX host, so we may need do at least the following enhancements: > 1) Enable live migration with even one nova compute. My current thinking is > that enhance target host as host:node when live migrate a VM instance and > the live migration task need > 2) Enable VCDriver report all ESX servers. > > We can discuss more during next week's IRC meeting. > > Thanks! > > > 2014-03-22 17:13 GMT+08:00 Shawn Hartsock <harts...@acm.org>: > >> Hi Jay. We usually use [vmware] to tag discussion of VMware things. I >> almost didn't see this message. >> >> In short, there is a plan and we're currently blocked because we have >> to address several other pressing issues in the driver before we can >> address this one. Part of this is due to the fact that we can't press >> harder on blueprints or changes to the VCDriver right now. >> >> I actually reported this bug and we've discussed this at >> https://wiki.openstack.org/wiki/Meetings/VMwareAPI the basic problem >> is that live-migration actually "works" but you can't presently >> formulate a command that activates the feature from the CLI under some >> configurations. That's because of the introduction of clusters in the >> VCDriver in Havana. >> >> To fix this, we have to come up with a way to target a host inside the >> cluster (as I pointed out in the bug) or we have to have some way for >> a live migration to occur between clusters and a way to validate that >> this can happen first. >> >> As for the priority of this bug, it's been set to Medium which puts it >> well behind many of the Critical or High tasks on our radar. As for >> fixing the bug, no new outward behaviors or API are going to be >> introduced and this was working at one point and now it's stopped. To >> call this a new feature seems a bit strange. >> >> So, moving forward... perhaps we need to re-evaluate the priority >> order on some of these things. I tabled Juno planning during the last >> VMwareAPI subteam meeting but I plan on starting the discussion next >> week. We have a priority order for blueprints that we set as a team >> and these are publicly recorded in our meeting logs and on the wiki. >> I'll try to do better advertising these things. You are of course >> invited... and yeah... if you're interested in what we're fixing next >> in the VCDriver that next IRC meeting is where we'll start the >> discussion. >> >> On Sat, Mar 22, 2014 at 1:18 AM, Jay Lau <jay.lau....@gmail.com> wrote: >> > Hi, >> > >> > Currently we cannot do live migration with VCDriver in nova, live >> > migration >> > is really an important feature, so any plan to fix this? >> > >> > I noticed that there is already bug tracing this but seems no progress >> > since >> > last year's November: https://bugs.launchpad.net/nova/+bug/1192192 >> > >> > Here just bring this problem up to see if there are any plan to fix >> > this. >> > After some investigation, I think that this might deserve to be a >> > blueprint >> > but not a bug. >> > >> > We may need to resolve issues for the following cases: >> > 1) How to live migration with only one nova compute? (one nova compute >> > can >> > manage multiple clusters and there can be multi hosts in one cluster) >> > 2) Support live migration between clusters >> > 3) Support live migration between resource pools >> > 4) Support live migration between hosts >> > 5) Support live migration between cluster and host >> > 6) Support live migration between cluster and resource pool >> > 7) Support live migration between resource pool and host >> > 8) Might be more cases..... >> > >> > Please show your comments if any and correct me if anything is not >> > correct. >> > >> > -- >> > Thanks, >> > >> > Jay >> > >> > _______________________________________________ >> > OpenStack-dev mailing list >> > OpenStack-dev@lists.openstack.org >> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev >> > >> >> >> >> -- >> # Shawn.Hartsock - twitter: @hartsock - plus.google.com/+ShawnHartsock >> >> _______________________________________________ >> OpenStack-dev mailing list >> OpenStack-dev@lists.openstack.org >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > > > -- > Thanks, > > Jay > > _______________________________________________ > OpenStack-dev mailing list > OpenStack-dev@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > -- # Shawn.Hartsock - twitter: @hartsock - plus.google.com/+ShawnHartsock _______________________________________________ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
