Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-01-05
14:39:36 mriedem kashyap was saying the libvirt devs were saying that qcow2 wouldn't work before libvirt 3.10 for shareable disks but raw should work, so this test run is setting CONF.libvirt.images_type=raw and CONF.use_cow_images=False
14:39:38 mriedem but still doesn't work
14:40:43 mnaser did i hit a really weird race condition possibly...? "Conflict updating instance e8cc3ace-2441-413e-bee7-d3a667e3298f. Expected: {'task_state': [u'migrating']}. Actual: {'task_state': None}"
14:41:13 mriedem possibly, might depend on what it's trying to change the task_state to
14:41:34 mnaser reset-state --active and doing a live-migration again triggers it.. let me check
14:41:59 Roamer` okay, so I'll try to also reach johnthetubaguy through a couple of other channels, but maybe I should also try to start a conversation here... What do people think about adding a pretty much trivial StorPool volume attachment driver as in https://review.openstack.org/140733/ ? I know that milestone 1 has passed and we're actually getting pretty close to milestone 2, but the Cinder and os-brick dri
14:42:05 Roamer` vers made it in last week, so right now we're in a funny situation where the main use case for the StorPool driver still needs manual patching. There's actually a major downstream OpenStack distributor with a client that wants to use StorPool, and the distributor would strongly prefer if all three StorPool drivers made it into OpenStack... so is there a chance for that? :)
14:42:14 mdbooth mriedem: I don't think those config params would affect volumes anyway fwiw.
14:43:06 mriedem was wondering that,
14:43:15 mriedem i guess those are only used for ephemeral imagebackend devices
14:43:21 mdbooth Yep
14:43:23 mriedem gah
14:43:23 mriedem ok
14:43:38 mriedem alright, trying again w/o pike uca so we use qemu 2.5
14:45:09 mriedem yikun: i think we should probably just add uuid to the response for the server migrations API in this change as well https://review.openstack.org/#/c/330406/ - is that something you think you'll have time for? if not, i can add that part of the change
14:50:58 mlavalle mriedem, hongbin: https://bugs.launchpad.net/neutron/+bug/1718605 is approved
14:50:59 openstack Launchpad bug 1718605 in neutron "[RFE] Support sub-string matching when filtering port with IP address" [Wishlist,Triaged] - Assigned to hongbin (hongbin034)
14:51:30 hongbin mlavalle: great. thanks
14:55:31 ildikov mriedem: got it, sadness, I guess it means we're falling back to play with the versions in Devstack then
14:56:28 mriedem ildikov: yeah
14:56:38 mriedem mlavalle: nice
14:57:59 fried_rice figleaf What's the status of the alternate hosts series (for purposes of the sched mtg agenda)?
14:58:06 fried_rice figleaf Are there open patches?
14:58:17 fried_rice figleaf Or just that bug?
15:01:02 mriedem fried_rice: i can tell you
15:01:11 fried_rice Tell me, tell me.
15:01:19 mriedem resize is the last piece https://review.openstack.org/#/c/526436/
15:01:29 mriedem but while figleaf was testing that, he found a race
15:01:35 mriedem so he's working on fixing that race
15:01:42 mriedem https://bugs.launchpad.net/nova/+bug/1741125
15:01:43 openstack Launchpad bug 1741125 in OpenStack Compute (nova) "Instance resize always fails when rescheduling" [High,Confirmed] - Assigned to Ed Leafe (ed-leafe)
15:03:21 fried_rice mriedem Thanks. Is the forked bug https://bugs.launchpad.net/nova/+bug/1741307 also related to alt hosts?
15:03:22 openstack Launchpad bug 1741307 in OpenStack Compute (nova) "Resize always fails when using the CachingScheduler" [High,In progress] - Assigned to Matt Riedemann (mriedem)
15:03:47 mriedem fried_rice: not to alternate hosts, just something i noticed when he pointed out the reschedule issue with resize
15:03:57 fried_rice cool, thanks mriedem
15:04:00 mriedem resize just doesn't work with the caching scheduler, period
15:04:03 melwitt mriedem: do you think this libvirt patch has had enough bake time to propose backports now? https://review.openstack.org/#/c/400384/
15:04:06 mriedem because of the migration allocation swap stuff
15:04:31 mriedem melwitt: heh, can you give it about another 18 months for us to know?
15:04:32 mriedem :)
15:04:41 melwitt D:
15:05:02 mriedem melwitt: honestly we probably won't know until people actually upgrade to start using that, and mnaser will probably be the first one once queens is released
15:05:17 mriedem melwitt: i'm fine with doing backports though
15:05:20 mriedem just not to newton
15:06:12 mriedem ildikov: the devstack patch is now just enablement https://review.openstack.org/#/c/266633/
15:06:18 mnaser i'll gladly take release notes for us :D
15:06:19 mriedem the patch on top of that turns on multiattach and disables the pike uca
15:07:18 melwitt mriedem: okay, cool. I shall propose them then. thanks
15:13:45 mnaser if there's any stable cores that can have a look at this backport: https://review.openstack.org/#/c/529385
15:17:40 openstackgerrit Lajos Katona proposed openstack/nova master: Deduplicate service status notification samples https://review.openstack.org/531381
15:17:41 mriedem dansmith: ^
15:17:49 dansmith mriedem: dun already got it
15:17:49 mriedem the stable thing, not the notificatoins thing
15:17:55 mriedem sympatico
15:17:56 dansmith bottom one needs a recheck?
15:18:10 lyarwood it already has
15:18:13 dansmith ah nm,
15:18:14 dansmith yeah
15:18:17 lyarwood should be in the gate by now
15:18:26 dansmith not quite
15:18:27 mriedem still check queue, but going
15:18:35 lyarwood urgh
15:18:52 mriedem dansmith: btw, i had to fix a test in https://review.openstack.org/#/c/530982/
15:19:34 mriedem lyarwood: since you're around https://review.openstack.org/#/q/I1aa3ca6cc70cef65d24dec1e7db9491c9b73f7ab
15:19:40 mriedem that should be the final thing for newton
15:20:04 mriedem melwitt: this is a simple bp driver feature parity change https://review.openstack.org/#/c/448034/
15:20:08 mriedem needs final +2
15:20:12 lyarwood mriedem: ack
15:20:36 melwitt mriedem: ack, will review
15:20:43 mriedem thanks
15:21:08 mriedem i see the allocation candidates limit change is also approved https://review.openstack.org/#/c/513526/ - nice, might get at least 2 bps merged today
15:21:28 openstackgerrit Lajos Katona proposed openstack/nova master: Deduplicate service status notification samples https://review.openstack.org/531381
15:22:35 mriedem andreykurilin: can you take a look at https://review.openstack.org/#/c/528128/ ? it will close out that bp and it's holding up other microversion changes in novaclient
15:27:18 ildikov mriedem: ok, tnx
15:27:34 openstackgerrit Lajos Katona proposed openstack/nova master: Deduplicate aggregate notification samples https://review.openstack.org/531162
15:30:43 mriedem stephenfin: http://logs.openstack.org/50/530950/1/check/legacy-tempest-dsvm-py35/f814425/logs/screen-n-novnc.txt.gz?level=TRACE#_Jan_04_17_19_48_242679
15:30:57 mriedem stephenfin: maybe try using encodeutils.safe_encode here? https://review.openstack.org/#/c/345399/37/nova/console/securityproxy/rfb.py
15:31:28 stephenfin mriedem: I can't catch a break with that thing :D That's a good idea though
15:31:51 stephenfin silly stephenfin
15:37:04 sean-k-mooney fried_rice: thanks. ya i noticed that there seemed to be multiple nrp patch series interwoven. i think https://review.openstack.org/#/c/384807/ is the patch i need to start using nrp to model hw acclerated vhost-vfio interfaces.
15:41:03 fried_rice sean-k-mooney "hw accelerated vhost-vfio interfaces". Sure.
15:42:25 mriedem i can't wait until sean-k-mooney stabs one of you guys
15:42:30 sean-k-mooney fried_rice: they are hardware offloaded port that use vfio-mdev instead of sriov to offload the dataplane and vhost to offload the control plane
15:43:18 fried_rice sean-k-mooney Okay, cool. So how are you planning to represent that from libvirt without using update_provider_tree?
15:43:32 hrw melwitt: thanks for +2 +W :)
15:45:36 sean-k-mooney fried_rice: step too is the libvirt xml generation but effectivly i am planning on having neutorn create teh resouce providers in placement and using a new vnic type to generate a request to placement for an allocation of the custom resouce class.
15:45:48 mriedem johnthetubaguy: is there some known thing with using the cachingscheduler where the compute services don't check in on time? or you have to adjust the compute service heartbeat?
15:46:08 johnthetubaguy mriedem: yeah, that is a thing
15:46:13 melwitt hrw: thanks for enhancing the commit message and the bug, makes it a lot easier to review
15:46:17 fried_rice sean-k-mooney This in Queens??
15:46:18 mriedem johnthetubaguy: orly
15:46:24 mriedem yeah because i'm seeing this in the scheduler logs
15:46:24 mriedem Jan 04 20:06:18.133057 ubuntu-xenial-rax-ord-0001690099 nova-scheduler[9487]: WARNING nova.scheduler.filters.compute_filter [None req-370806fa-5640-485d-adc2-21808de9ccb4 tempest-MigrationsAdminTest-1101936831 tempest-MigrationsAdminTest-1101936831] (ubuntu-xenial-rax-ord-0001690099, ubuntu-xenial-rax-ord-0001690099) ram: 6446MB disk: 58368MB io_ops: 15 instances: 15 has not been heard from in a while
15:46:28 mriedem but no errors in the n-cpu logs
15:46:31 hrw melwitt: o yes ;D
15:46:46 hrw melwitt: can you take a look at https://review.openstack.org/#/c/530965/ too?
15:46:46 johnthetubaguy mriedem: if your cache interval is larger than than how long you thing it is dead, it starts to ignore some of your nodes randomly
15:46:54 mriedem ah
15:46:54 hrw melwitt: I am open for any questions
15:46:57 mnaser mriedem: i think i found out how i got that weird state. doing a host-evacuate-live sends a live-migration request for all instances, as nova processes them one by one, the states of all vms in the compute host are in "migrating".. but then its possible that a delete request comes in while the live migration is happening.

Earlier   Later