Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-08-08
16:25:35 jaypipes dansmith: well, this method will need to re-create allocations that an ocata compute node deleted incorrectly.
16:26:20 jaypipes dansmith: if an ocata compute node was the source for a migrate to a pike compute host.
16:26:31 dansmith but that's already done in update_usage_for_instance right?
16:26:46 dansmith in the top patch, conditional on the presence of ocata computes
16:28:14 jaypipes dansmith: ack, sorry, yes, you're right.
16:28:22 jaypipes dansmith: I can rename it back to delete...
16:28:32 dansmith jaypipes: I'm not sure what the point of this patch is though.. just some extra conditions for not calling delete?
16:29:51 jaypipes dansmith: the point of this patch was two things: 1) to only delete allocations when the instance is DELETED or SHELEVED_OFFLOADED and 2) add a bunch of log statements so we can see what state instances are in during this _remove_deleted_instances_allocations() method
16:30:44 dansmith jaypipes: okay I thought we were doing this filtering earlier such that by the time we got here we didn't have anything other than the ones we needed to delete..
16:30:55 jaypipes dansmith: I added the debug log statements when trying to fix the issue alex_xu brought up about evacuate code paths not being correct.
16:31:09 dansmith we've been round and round about this particular conditional, so maybe I missed something
16:31:33 jaypipes dansmith: the evacuate code paths were not being affected by the update_usage_for_instance() path for some reason.
16:31:45 jaypipes dansmith: on the source host at least.
16:32:34 jaypipes dansmith: so when trying to verify the evacuate code path for scenario when source host is restarted, I was trying to get log messages about the instances in this method. sorry, it got really confusing :(
16:32:55 dansmith jaypipes: so vm_states.ALLOW_RESOURCE_REMOVAL excludes things that aren't deleted or shelved_offloaded, so doesn't that cover us?
16:33:07 dansmith oh, not for deleting allocations I guess
16:33:17 jaypipes right, that actually results in InstanceNotFound.
16:33:27 dansmith what does?
16:33:32 jaypipes when an instance is deleted.
16:33:41 jaypipes the Instance.get_by_uuid() returns InstanceNotFound
16:34:47 openstackgerrit Stephen Finucane proposed openstack/nova master: doc: Import configuration reference https://review.openstack.org/491853
16:35:04 mriedem stephenfin: in case you haven't left yet, there are two duplicate upgrade notes for similar things in the libvirt driver about port filtering capability
16:35:04 mriedem https://docs.openstack.org/releasenotes/nova/unreleased.html#id15
16:35:05 jaypipes dansmith: sorry about all of this :( in trying to address alex_xu's review comments about evacuate I ended up going down the rabbithole.
16:35:21 jaypipes dansmith: and it's likely I've f'd it all up now :(
16:35:25 dansmith jaypipes: okay, yeah, I'm confused I guess
16:35:32 stephenfin mriedem: I assume we can just delete one/
16:35:34 stephenfin *?
16:35:34 dansmith jaypipes: I'm throwing some comments in here though
16:35:40 jaypipes k
16:35:45 mriedem stephenfin: they are slightly different
16:36:34 mriedem https://github.com/openstack/nova/blob/501a0fe65afabb7e6a8b1e2d1ca0fbcea941d2d1/releasenotes/notes/libvirt-ignore-allow_same_net_traffic-fd88bb2801b81561.yaml
16:36:38 mriedem https://github.com/openstack/nova/blob/5c0cb33d75acb284a3b6299c851ab70f10f69dfc/releasenotes/notes/libvirt-firewall-ignore-use_ipv6-c555f95799f991fd.yaml
16:37:31 openstackgerrit Chris Friesen proposed openstack/nova master: Remove ram/disk sched filters from default list https://review.openstack.org/491854
16:40:24 mriedem stephenfin: maybe i'll just make sure the boilerplate text is the same between the two
16:40:27 mriedem and leave the rest
16:41:06 stephenfin mriedem: As in keep both but make the shared part look similar?
16:41:14 cfriesen mriedem: dansmith: how much validation of images should nova do? I'm getting a report that with RDB instance storage launching an instance from qcow2 format image doesn't boot but launching instance from raw format image works. Booting from volume with either image works.
16:41:18 stephenfin or something else?
16:41:22 mriedem stephenfin: yes
16:41:38 stephenfin mriedem: Cool. You or me doing that?
16:42:06 mriedem i'm doing it
16:42:49 stephenfin (y)
16:43:12 cfriesen mriedem: dansmith: similarly, should we believe glance when it tells us the image type? In Newton at least glance doesn't appear to be validating the specified image type.
16:44:42 openstackgerrit Matt Riedemann proposed openstack/nova master: Cleanup release note about ignoring allow_same_net_traffic https://review.openstack.org/491855
16:44:45 bauzas folks, bailing out, see you Thursday \o
16:50:52 stephenfin bauzas: o/
16:52:01 stephenfin mriedem: +2d
16:52:12 stephenfin sdague: email sent
16:52:17 stephenfin (I hope - Evolution sux)
16:55:07 openstackgerrit Stephen Finucane proposed openstack/nova master: doc: Import configuration reference https://review.openstack.org/491853
16:57:08 stephenfin See you all next week o/
17:04:09 cdent jaypipes, bauzas, dansmith, mriedem : I made a new patch to publish the placement-api-ref. https://review.openstack.org/#/c/491860/ it depends on the top patch in https://review.openstack.org/#/q/topic:cd/placement-api-ref+status:open so it would be nice if we could get those in and have happy little docs
17:04:27 jaypipes cdent: kk
17:17:51 melwitt dansmith, mriedem: my bfv functional test queries hypervisor stats to verify correct handling of bfv. it seems like that's independent of the RT, right?
17:18:08 dansmith no, that's where hv stats come from
17:19:10 melwitt so currently the scheduler, claims, and reporting will be different than hv stats?
17:19:44 melwitt that is, they'll show different resource amounts?
17:19:45 openstackgerrit Doug Hellmann proposed openstack/nova master: use intersphinx to link to horizon https://review.openstack.org/491866
17:20:42 dansmith melwitt: we do account for things a little differently in placement, which is what scheduler is using for the basis of choosing a host, and claiming against
17:21:01 dansmith hv stats has never been particularly correct either, mind you
17:21:36 melwitt okay
17:22:17 melwitt I think I could update the test to query placement for how much disk the compute node is reporting
17:22:43 melwitt because that's all I was doing is verifying it's reporting 0 local_gb_used
17:23:32 dansmith so the compute node's allocation will account for it, but we don't heal if we're on pike only,
17:23:40 melwitt it would be interested to see if the test passes with those lines commented out too, because if bfv isn't handled correctly yet, it won't even be able to boot instances
17:23:50 melwitt *interesting
17:24:03 dansmith looking earlier, the place I had added it to the scheduler request didn't seem particularly correct anymore
17:24:06 sean-k-mooney hi everyone quick question is "nova-manage cell_v2 create_cell" updating the db directly or calling the api to create the cell entries?
17:24:15 dansmith sean-k-mooney: directly
17:24:24 dansmith sean-k-mooney: almost all of nova-manage is direct-to-databsae
17:24:54 sean-k-mooney dansmith: cool thanks, that is what i taught. trying debug an issue in kolla and wnated to check where the logs would be
17:25:26 dansmith on the bfv thing, I'm a little concerned about even getting jaypipes' set in before rc at this point, so...
17:28:44 melwitt yeah. I've been trying to mention the bfv thing occasionally throughout this cycle because it's one of the top things people are looking for from resource providers and I haven't been able to tell how close we are to that working
17:29:30 melwitt I'll try stacking the test on top of jay's set locally to see if it passes without the hv stats stuff
17:30:24 dansmith jaypipes' set doesn't affect any of this, AFAIK
17:30:52 dansmith and we're not even accounting for moves properly right now (which is the subject of the set) which is arguably more important
17:32:05 dansmith my zero request patch may actually still work, since we're still calling resources_from_request_spec or whatever it's calls
17:32:05 cdent dansmith: have you got a summary of “what’s wrong today”? It’s hard to keep track.
17:32:07 dansmith *called
17:32:18 dansmith cdent: no
17:32:20 openstackgerrit Doug Hellmann proposed openstack/nova master: use extlinks to build series-specific links https://review.openstack.org/491869
17:32:43 cdent roger that, I’ll travel with lights
17:33:39 melwitt dansmith: okay. I thought earlier you said we're already accounting for it in the current stuff, I must have misunderstood
17:34:05 dansmith melwitt: I said we are on the compute node, but that the scheduler probably isn't (which I since confirmed)
17:34:15 dansmith and if we're on pike we'll depend on the scheduler having it right
17:34:26 melwitt I see, thanks
17:34:36 dansmith which is what my zero request patch is for.. asking the scheduler for zero root disk
17:34:38 openstackgerrit Spencer Yu proposed openstack/python-novaclient master: Nova client should retry with Retry-After value https://review.openstack.org/447766
17:34:50 melwitt yeah
17:35:11 dansmith this is somewhat wrapped up in the shared storage thing which we're punting on again, so it really can't be right regardless
17:36:39 melwitt okay. yeah, I was wondering about the shared storage thing too. good to know
17:38:49 jaypipes melwitt: yeah, the bfv accounting issues will most likely remain in Pike unfortunately.
17:42:27 dansmith jaypipes: replied
17:43:15 dansmith jaypipes: I still don't see how we get to the offload thing, and just want that code to be less complicated, but if we're not sure or you've seen it in real life somehow, then we should just leave it in I guess
17:44:50 jaypipes dansmith: this patch is mostly in response to mriedem and alex_xu's comments on https://review.openstack.org/#/c/488510/ ps 27.
17:45:12 jaypipes dansmith: i.e. this from mriedem:
17:45:12 jaypipes Related to #4, there is bug https://bugs.launchpad.net/nova/+bug/1679750 where we don't delete allocations from Placement in nova-api when doing a "local delete", i.e. when the compute host is down, or the instance does not have a host associated (e.g. shelved_offloaded state).
17:45:13 openstack Launchpad bug 1679750 in OpenStack Compute (nova) "Allocations are not cleaned up in placement for instance 'local delete' case" [Medium,In progress]
17:45:43 jaypipes dansmith: but I admit I've probably f'd this series up :(

Earlier   Later