Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-11-01
14:55:53 dansmith well, because it moved
14:56:08 dansmith anyway, that's a door we don't want to open, even under config I think
14:56:08 mriedem yeah
14:56:32 dansmith deleting only if we find an actually-deleted instance would be one way to handle it,
14:56:43 dansmith which pushes the race to the delete-archive boundary
14:56:48 dansmith but that's much better, IMHO
14:57:17 dansmith if they're doing fast purging (like immediate purging of all deleted things, then they could bump into it, but much less likely, and we could have nova-manage check for allocations potentially
15:01:54 openstackgerrit Matthew Booth proposed openstack/nova master: libvirt: Don't VIR_MIGRATE_NON_SHARED_INC without migrate_disks https://review.openstack.org/507202
15:02:20 openstackgerrit Matthew Booth proposed openstack/nova master: DNM: Run test_volume_backed_live_migration and iscsi test https://review.openstack.org/508163
15:03:31 openstackgerrit Jianghua Wang proposed openstack/nova master: XenAPI: get vGPU stats from hypervisor https://review.openstack.org/512965
15:05:16 jianghuaw jaypipes, thanks for the comment. Yes, you're correct. The above is the reworked revision.
15:06:20 openstackgerrit OpenStack Proposal Bot proposed openstack/os-vif master: Updated from global requirements https://review.openstack.org/511035
15:06:22 mriedem dansmith: ok here is the bug https://bugs.launchpad.net/nova/+bug/1729371
15:06:23 openstack Launchpad bug 1729371 in OpenStack Compute (nova) "ResourceTracker races to delete instance allocations before instance is mapped to a cell" [High,Triaged]
15:06:41 dansmith mriedem: roger
15:10:35 mriedem stephenfin: can you take a look at https://review.openstack.org/#/c/481116/ and below?
15:10:52 mriedem i'd like to get that fixed and backported through all supported stable branches since it's a regression since newton
15:11:13 mriedem well, less of a regression than a busted feature since newton for evacuating with a target host
15:12:59 openstackgerrit Matthew Booth proposed openstack/nova master: Remove unused block_migration argument to _live_migration_operation https://review.openstack.org/517007
15:15:21 jianghuaw dansmith, may you take a look at this patch adding an option for enabled_vgpu_type?
15:15:23 jianghuaw https://review.openstack.org/#/c/512580/
15:17:33 dansmith jianghuaw: I have a short day today and a lot in the queue, but I will try
15:19:23 jianghuaw dansmith, understood. thanks.
15:19:53 openstackgerrit Dan Smith proposed openstack/nova master: WIP Avoid deleting allocations for instances being built https://review.openstack.org/517009
15:20:11 dansmith mriedem: this is what I'm thinking ^
15:21:32 mriedem dansmith: yeah i like that
15:21:47 mriedem still maintains the InstanceNotFound logic if the instance is in the cell db and deleted
15:21:56 dansmith we log the situation so if they see the same instance over and over, they know some allocation is stuck
15:22:20 dansmith we _could_ track that set run-to-run and log more forcefully, but that's getting more complex
15:22:43 mriedem https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcQ1b-BPsqPMmmI9aPgemul62fGgzoXzhuXWTEde_tBvLbcFaRAA
15:22:52 mriedem always good to know about the situation
15:23:01 dansmith um
15:24:29 mriedem if you want i can handle updating tests and such for that if you're busy today
15:25:22 dansmith I'm working on it now
15:25:28 dansmith I'll punt if I need to
15:30:08 mriedem anthonyper: fyi https://bugs.launchpad.net/nova/+bug/1728924
15:30:09 openstack Launchpad bug 1728924 in OpenStack Compute (nova) "console logging does not work for OL instances on xen compute" [Undecided,Confirmed]
15:32:28 dansmith mriedem: heh, we have a test that asserts that we abort these unborn children
15:32:55 mriedem i'm not sure how to appropriately respond to that statement
15:33:17 dansmith it asserts the desired behavior for reasons other than "every instance goes through this state" of course
15:33:26 dansmith but it's basically asserting that this buggy behavior exists
15:33:52 mriedem huh
15:34:00 mriedem unrelated, i suppose bauzas is in transit
15:34:12 dansmith already?
15:34:16 mriedem jaypipes: just saw another bug in triage for this same NFS resize bfv issue https://review.openstack.org/#/c/516395/
15:34:20 mriedem we should get that fixed and backported
15:34:26 mriedem s/fixed/merged/
15:34:38 mriedem note that https://review.openstack.org/#/c/516396/ verifies it
15:34:41 mriedem melwitt: you too ^
15:34:46 mriedem dansmith: only assuming
15:34:53 mriedem i know he likes to fly out about a week early :)
15:35:27 dansmith lol right
15:42:26 Nisha_Agarwal jaypipes, hi
15:42:47 Nisha_Agarwal jaypipes, i had a query regarding traits scheduling
15:44:34 Nisha_Agarwal jaypipes, is there any prefernce associated with scheduling? among traits scheduling and capabilities scheduling?
15:45:13 Nisha_Agarwal mriedem, ^^^
15:45:50 mriedem idk
15:45:58 Nisha_Agarwal johnthetubaguy, ^^^^ do u have any idea?
15:47:24 dansmith Nisha_Agarwal: what does that mean "capabilities and traits"
15:47:56 dansmith placement will consider resource amounts and traits, the scheduler will run filters on the results based on what we have today
15:49:00 Nisha_Agarwal dansmith, what do you mean by " on what we have today"?
15:49:22 Nisha_Agarwal dansmith, scheduler runs the filters after placement is executed?
15:50:11 dansmith Nisha_Agarwal: yes
15:50:39 Nisha_Agarwal dansmith, i was asking from the point that if a node has both traits and capabilities populated which one gets the preference in scheduling...is there any preference associated with it?
15:51:01 dansmith Nisha_Agarwal: and I'm saying I don't know what "capabilities" are in this context
15:51:10 mriedem dansmith: i think the CapabilitiesFilter
15:51:21 Nisha_Agarwal mriedem, yes
15:51:22 dansmith so in that case, traits win
15:51:27 dansmith as I said,
15:51:45 dansmith we select based on resource counts and traits, then apply the filters on the result
15:51:54 dansmith if we have a filter, then that will filter the result set
15:54:19 Nisha_Agarwal dansmith, but there can be a case where say the traits scheduling selects 0 nodes but capabilitiesfilter would have selected 1 node if it was run on the same original set of nodes
15:54:44 dansmith yup
15:55:47 Nisha_Agarwal dansmith, then in that case when scheduling done along with traits+capabilities, and capabilities alone differ. :(
15:56:32 dansmith Nisha_Agarwal: agree, except with the sadface
15:57:33 dansmith Nisha_Agarwal: our goal is to make scheduling better, not to preserve every detail of the existing behavior forever
15:57:41 openstackgerrit Dan Smith proposed openstack/nova master: Avoid deleting allocations for instances being built https://review.openstack.org/517009
15:57:52 dansmith mriedem: ^
15:57:55 openstackgerrit Hamdy Khader proposed openstack/nova master: Adding NVMEoF for libvirt driver https://review.openstack.org/482640
15:58:12 dansmith mriedem: lots of test noise in there because of how those tests work, unfortunately
16:00:21 mriedem ok
16:11:29 jaypipes mriedem: k, will look shortly.
16:11:39 jaypipes Nisha_Agarwal: just got back to the office... gimme a few
16:13:10 openstackgerrit Jay Pipes proposed openstack/nova master: required traits for no sharing providers https://review.openstack.org/517027
16:13:21 jaypipes efried, alex_xu: ^
16:13:48 efried jaypipes Ack. I did some refactoring of the bottom patch - do you have local deltas?
16:14:51 jaypipes efried: yes
16:15:01 jaypipes efried: I pulled the latest from this morning.
16:16:16 efried jaypipes k, I'll wait. My deltas are improving the readability of that method. Only one of my tweaks actually changes the SQL that's produced, but the test result is still good, so if that's gonna break something, I'll need a test for it.
16:16:34 efried jaypipes We can discuss this p.m. if you're around.
16:17:10 jaypipes efried: readability of the test method or readability of the _get_provider_ids_no_share() method?
16:17:29 efried jaypipes Of _get_provider_ids_matching_all
16:18:10 efried jaypipes Mainly renaming variables and getting rid of unnecessary ones. But the SQL-changer is that 'jc' doesn't need to be linked to the lastij - it can always just compare against rpt.c.id.
16:18:30 jaypipes efried: ah, sure, ok.
16:18:35 efried jaypipes If that's not true, then I need a test path that brings it out.
16:18:49 efried But I don't see why it shouldn't be okay.
16:18:59 jaypipes efried: yes, like I said, I copied the code from the get_all_with_shared() and just removed things. didn't change anything. so I'm good with a change like that, sure.
16:19:19 efried jaypipes Okay, cool. I'll post it once you've done yours.
16:19:30 efried jaypipes Because I'm all about rebase, 'bout rebase. No trouble.
16:20:08 jaypipes efried: why don't you do that now. I can always stash, stash pop after pulling your latest.

Earlier   Later