Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-07-14
08:02:01 Habib-A Hello all please help :D https://ask.openstack.org/en/question/128468/windows-server-2019-with-ovs-and-nova-compute/
08:26:59 gibi kashyap: hi! you might have knowledge that I could use. I see a strange behavior from libvirt / qemu . with libvirt 4.0.0 and qemu 2.11 I can only hotplug an SRIOV VF to a guest if I detach the VF first from the host even if I specify managed=yes in the xml. The domain create works, libvirt automatically detach the device from the host.
08:27:18 kashyap gibi: Mornin
08:27:35 gibi kashyap: then I upgraded to libvirt 6.0.0 and qemu 4.2 and now hotplug also works without a manual detach
08:27:56 gibi I looked at the changelog of libvirt but nothing pops up
08:28:35 kashyap gibi: So a "feature was introduced silently"? :-)
08:28:58 gibi I think it is more like bugfix than a feature
08:29:18 kashyap gibi: A question: is the libvirt-6.0.0 & qemu-4.2 behaviour desirable to you, or not? (Let's keep the surprise part aside :))
08:29:27 kashyap Right, I first wrote 'bug-fix' indeed
08:29:27 gibi desirable
08:29:35 gibi very mutch
08:29:55 gibi I'm working on nova to support sriov attach
08:30:10 gibi and for that this behavior is needed
08:30:25 kashyap gibi: So you want to know whether this is intended or not?
08:30:45 gibi I want to know what is the minimum libvirt / qemu version that this behavior exists
08:31:56 kashyap gibi: Ah, good one; I can check w/ one of the upstream libvirt folks
08:32:04 gibi kashyap: thanks a lot!
08:32:26 kashyap gibi: I'll get back to you in a bit. Today also I need to finish some administrative paperwork (deadline is 16 Jul; dammit). But it's on my to-do
08:32:41 gibi kashyap: thanks. it is not urgent
08:33:10 kashyap gibi: But yeah; silently introducing such a user-visible bug fix without even a changelog is not nice
08:38:30 gibi I'm glad they fixed it :)
09:09:15 openstackgerrit Stephen Finucane proposed openstack/nova master: trivial: Use constants for libvirt version checks https://review.opendev.org/740921
09:09:15 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove support for Intel CMT events https://review.opendev.org/740922
09:33:17 stephenfin gibi: You're working on SR-IOV attach, you say?
09:33:50 stephenfin That's interesting, because that's on the backlogs of sean-k-mooney and I for quite some time now, and we simply hadn't gotten to it yet
09:37:41 openstackgerrit Huaqiang Wang proposed openstack/nova master: objects: Introduce the 'CPUAllocationPolicy.MIXED' enum https://review.opendev.org/716267
09:37:41 openstackgerrit Huaqiang Wang proposed openstack/nova master: hardware: refactor the code for CPU policy sanity check https://review.opendev.org/740925
09:40:23 openstackgerrit Huaqiang Wang proposed openstack/nova master: hardware: refactor the code for CPU policy sanity check https://review.opendev.org/740925
09:40:51 stephenfin gibi: I parroted your question back to danpb, given kashyap was occupied - as far as he's concerned, the behaviour you saw on the older version is bogus and reflects a bug
09:41:04 stephenfin "as managed=yes stuff has been around forever and never required a manual detach"
09:41:35 kashyap sfinucan: Thanks; you checked off a TODO on my list :)
10:12:49 openstackgerrit Huaqiang Wang proposed openstack/nova master: [DNM] bump nova-compute version for mixed instance in API https://review.opendev.org/740933
10:14:23 huaqiang stephenfin: I have sent out another ongoing change as a possible alternative of https://review.opendev.org/716271
10:14:51 huaqiang https://review.opendev.org/716271 checks the nova-compute service version in conductor layer
10:15:17 huaqiang while https://review.opendev.org/740933 is doing this in API layer
10:15:26 huaqiang just for comparison
10:19:32 gibi stephenfin: yeah, I did not assigned the bp or file a spec for it as I did not know how much time I will have during this cycle. So basically I spent yesterday putting a PoC together that started working today morning. I will do the paperwork at some point before M2
10:20:22 gibi regarding the libvirt bug, it would be nice to know what is the minimum libvirt version that works for documentation purposes
10:28:29 stephenfin huaqiang: ack, will look
10:29:11 stephenfin gibi: Okay, cool. I know Sean and I would be keen to take a look at that, so if/when you do lemme know
10:30:40 stephenfin I can't help with testing at the moment since my SR-IOV machines are at home (I'm staying with my folks in the countryside since March), but hopefully it was just a single bad version of libvirt. Can probably test by building various versions from source
10:30:59 stephenfin Once I get access to those machines, I could do that testing
10:39:01 gibi stephenfin: OK, thanks. I have access to an SRIOV capable machine so I can try different libvirt version too
10:50:22 gibi I think I will be able to post some WIP code this week
11:20:19 stephenfin huaqiang: Okay, I prefer https://review.opendev.org/#/c/740933/
11:21:09 stephenfin May be wise to get someone that knows RPC better than I do (dansmith or gibi, perhaps?) to weigh in
11:25:53 stephenfin lyarwood: I had my message asking you to look at the vTPM work already typed out before I checked Gerrit 😅 Thanks for the reviews
11:25:53 stephenfin lyarwood: I had my message asking you to look at the vTPM work already typed out before I checked Gerrit 😅 Thanks for the reviews
11:32:46 gibi stephenfin, huaqiang: looking at the alternatives I think I would go with the check in the api service
11:33:33 stephenfin gibi: so you mean prefer https://review.opendev.org/#/c/740933/ over https://review.opendev.org/716271 ?
11:34:05 gibi yes I prefer https://review.opendev.org/#/c/740933/
11:37:08 gibi it is a bit more explicit what operations we block if the check is in the API
11:38:08 gibi and also understanding the code is easier for me if the check is not burried down in a long call path
11:39:50 openstackgerrit Merged openstack/nova stable/stein: Check cherry-pick hashes in pep8 tox target https://review.opendev.org/737922
11:41:49 openstackgerrit Jinsheng Zhang proposed openstack/nova-specs master: Add nova-support-multiple-boot-volume-with-boot-order-selection spec https://review.opendev.org/737897
11:45:35 huaqiang stephenfin and gibi, thanks for comments!
11:59:45 openstackgerrit Stephen Finucane proposed openstack/nova master: WIP: Attempt to restore some sanity to snapshot https://review.opendev.org/740336
11:59:45 openstackgerrit Stephen Finucane proposed openstack/nova master: libvirt: Use better variable names for '_create_guest' https://review.opendev.org/740945
12:09:49 stephenfin lyarwood: replied at https://review.opendev.org/#/c/739209/4
12:21:01 lyarwood stephenfin: just back sorry
12:21:04 lyarwood stephenfin: ack thanks
12:55:07 openstackgerrit Lee Yarwood proposed openstack/nova stable/train: Silence amqp heartbeat warning https://review.opendev.org/728057
12:57:23 lyarwood melwitt: https://review.opendev.org/#/c/732708/ - would you mind hitting this once you're online today, the stable/stein change has merged.
12:57:40 lyarwood and the func test below it please :)
14:06:49 mnaser so neutron makes 0 use of wrap_db_retry and has ssome convluted 4 year old retry mechanism which tries to avoid nested use of wrap_db_retry (well, if it detects it, it removes it, sounds like a workaround to me)
14:06:58 mnaser does nova have some policy on what is retried and what isn't?
14:07:44 dansmith mnaser: we decorate the methods we want to retry, IIRC
14:08:03 mnaser dansmith: ok so there isn't a "if X then decorate" policy that i can try and adopt for neutron's db stuff
14:08:19 dansmith mnaser: I think we also don't have a lot of nesting of things that might be called from the top, or called from within, so no nesting
14:08:29 dansmith mnaser: yes, the policy is "if we get a bug for a thing, decorate it" :)
14:08:55 mnaser ok, so the current what-seems-to-be-neutron policy is... decorate everything with a decorator that removes an extra one if it finds one to avoid nested retries
14:09:05 mnaser so this is going to be very fun to unwind
14:56:32 dansmith gmann: tres woot: https://review.opendev.org/#/c/734184/
14:57:58 gmann dansmith: nice, checking
14:58:26 dansmith (moved to -qa)
16:09:50 openstackgerrit Balazs Gibizer proposed openstack/nova master: [WIP] Support SRIOV interface attach and detach https://review.opendev.org/740995
16:10:44 gibi stephenfin, sean-k-mooney: this might be intresting for you ^^
16:12:10 openstackgerrit Dan Smith proposed openstack/nova master: Make our ceph job test with glance in multistore mode https://review.opendev.org/734184
16:42:31 openstackgerrit Huaqiang Wang proposed openstack/nova master: notifications: add the pcpuset info to instance numa cell payload https://review.opendev.org/728480
16:42:31 openstackgerrit Huaqiang Wang proposed openstack/nova master: objects: Introduce the 'CPUAllocationPolicy.MIXED' enum https://review.opendev.org/716267
16:42:32 openstackgerrit Huaqiang Wang proposed openstack/nova master: hardware: refactor the code for CPU policy sanity check https://review.opendev.org/740925
16:42:32 openstackgerrit Huaqiang Wang proposed openstack/nova master: libvirt: apply mixed instance CPU policy https://review.opendev.org/728481
17:08:01 openstackgerrit Huaqiang Wang proposed openstack/nova master: Calculate the CPU usage for mixed instance https://review.opendev.org/716269
17:08:02 openstackgerrit Huaqiang Wang proposed openstack/nova master: scheduler: mixed instance CPU resource translating https://review.opendev.org/728482
17:25:55 openstackgerrit Huaqiang Wang proposed openstack/nova master: Validate CPU pinning configuration for mixed instance https://review.opendev.org/716268
17:25:56 openstackgerrit Huaqiang Wang proposed openstack/nova master: Calculate the CPU usage for mixed instance https://review.opendev.org/716269
17:25:56 openstackgerrit Huaqiang Wang proposed openstack/nova master: scheduler: mixed instance CPU resource translating https://review.opendev.org/728482
18:53:20 openstackgerrit Merged openstack/nova master: utils: Move 'get_bdm_image_metadata' to nova.block_device https://review.opendev.org/739996
20:30:54 dansmith sean-k-mooney: replied. thanks for reviewing and for your initial help in that of course
20:33:00 sean-k-mooney no worries
20:36:34 sean-k-mooney dansmith: ya makes sense to me
20:36:55 sean-k-mooney im glad you finally got it working by the way.
20:37:06 sean-k-mooney its always nice when the ci verifys something like this for us
20:37:47 sean-k-mooney the glance folks might even want to take a look for there own testing
21:36:25 openstackgerrit Merged openstack/nova stable/rocky: Add functional test for bug 1550919 https://review.opendev.org/732707
21:36:25 openstack bug 1550919 in OpenStack Compute (nova) rocky "[Libvirt]Evacuate fail may cause disk image be deleted" [Medium,In progress] https://launchpad.net/bugs/1550919 - Assigned to Lee Yarwood (lyarwood)
21:54:44 dansmith sean-k-mooney: yeah, they'll have a lot more surface they can test once my two devstack patches land.. I hope they will make use of that, but.. who knows
22:24:03 sean-k-mooney dansmith: well if we are testing it its better then nothing
23:20:46 openstackgerrit Merged openstack/nova stable/rocky: libvirt: Don't delete disks on shared storage during evacuate https://review.opendev.org/732708
23:37:53 mnaser melwitt: i know you've worked some with the quota code. does nova have any locking on quota usages? neutron currently has a row-level lock on a resource/tenant_id which is causing me worlds of pain on highly parallelized port creation

Earlier   Later