Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-10-15
13:09:31 sean-k-mooney so if you have 1 free compute node you can live mgirate without thse backports to do a rolling upgrade
13:10:10 sean-k-mooney you just need to ensure the cpus and hugepages exctra are free on the dest
13:10:27 brinzhang yes
13:11:40 brinzhang I will use two node to have a test, at leat two node, because I should ensure this feature is fine, otherwise, it will have some risk, it's not allow
13:12:33 brinzhang Later I will search for related bug fixes and backport
13:13:00 sean-k-mooney have you also backported sriov live migration i assume not
13:13:55 brinzhang sriov live migration we completed it by ourself, dont need to backport ^^
13:14:34 brinzhang If we upgrade in future, and we will use the merged codes too
13:14:45 sean-k-mooney the reason i bring that up is numa migratio nmerged after sriov and at least 1 of the patches in the numa serises is adding compatiablty with sriov live migrfation
13:15:07 sean-k-mooney your going from rocky to train right
13:15:19 brinzhang yes, it contained, I was found your point
13:15:38 sean-k-mooney train is the verssion that had both so once your on train you shoudl be fine
13:15:41 brinzhang maybe from rocky to ussuri
13:16:27 brinzhang because we also want to use the pmem feature too ^
13:17:16 brinzhang sean-k-mooney: thanks for your help
13:17:43 brinzhang I will go home now, have a holiday sean-k-mooney, gibi :D
13:17:56 gibi brinzhang: o/
13:41:03 dansmith sean-k-mooney: compute=auto will try to select the version supported by the oldest compute service in the system, based on service_version
13:41:43 sean-k-mooney dansmith:yes
13:41:53 sean-k-mooney the max version suspport by the oldest
13:42:03 sean-k-mooney which is why i was saying to use 5.1
13:42:07 sean-k-mooney not auto
13:42:26 sean-k-mooney since brinzhang is not update all the nodes
13:42:29 dansmith ah, you said [05:56:27] <sean-k-mooney>so i think that will default to the newest version supported by the current node
13:42:38 dansmith but you meant that for if it's unset or something?
13:42:48 sean-k-mooney yes unset
13:43:06 dansmith okay I was skimming the backscroll and thought you were saying that was the auto behavior,
13:43:08 dansmith but I see now
13:58:03 bbowen sean-k-mooney: Do you know the latest on https://trello.com/c/f0lLrWxf/2465-swisscom-schweiz-agsev302710209bz1860395neutronrarp-packets-after-live-migration-are-dropped-by-br-link-due-to-a-timing-issue-be, I thought you were making a change and it was not on neutron team. was a hint in rhos-prio to go back to compute
13:59:06 bbowen ^now on neutron
13:59:18 lyarwood bbowen: wrong channel :)
13:59:44 bbowen ty
14:34:51 artom stephenfin, I think I'd like a video chat to explain https://review.opendev.org/#/c/756070/6/nova/virt/hardware.py if you have some time today or next week
14:34:59 artom Not so much the change itself, more the context around it
14:44:24 artom You set `instance_cell.min_threads = threads_no`, and I'm trying to work backwards to where threads_no is coming from, and it's... hard :P
15:07:56 openstackgerrit Balazs Gibizer proposed openstack/nova master: Use the non polling notification waiter in func test https://review.opendev.org/758445
15:07:57 openstackgerrit Balazs Gibizer proposed openstack/nova master: Create a fixture around fake_notifier https://review.opendev.org/758446
15:09:18 openstackgerrit Balazs Gibizer proposed openstack/nova master: Use NotificationFixture for legacy notifications too https://review.opendev.org/758448
15:10:38 openstackgerrit Balazs Gibizer proposed openstack/nova master: Test the NotificationFixture https://review.opendev.org/758450
15:10:39 openstackgerrit Balazs Gibizer proposed openstack/nova master: Move fake_notifier impl under NotificationFixture https://review.opendev.org/758451
15:13:21 gmann gibi: I have doctor appointment and would not be able to join today meeting.
15:13:35 gibi gmann: ack, thanks for the headsup
15:42:23 lyarwood stephenfin: you've been working on LM bugs recently with the periodic jobs right?
15:42:53 lyarwood stephenfin: did you find any issues with _sync_instance_power_state running *after* an instance had already paused?
15:49:46 artom lyarwood, I think he's on PTO this EU afernoon - he cancelled a meeting we had with Brian and Erwan
15:49:56 lyarwood ah forgot my bad
15:49:59 lyarwood thanks artom
15:51:07 gibi nova meeting starts in 10 minutes on #openstack-meeting-3
16:13:48 dansmith melwitt: ahh, maybe the separate joins are only happening when we grab multiple instances
16:14:12 dansmith if that's the case, then just making the get-by-uuid path work the same would be easy
16:15:23 melwitt is there a specific part of the code you're looking at that makes you think that?
16:15:37 dansmith I shouldn't say "separate joins"... they're called "manual joins" in the db api, because they're separate queries
16:15:56 dansmith melwitt: look at this: https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L1278
16:16:05 dansmith melwitt: and this: https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L1328
16:16:25 melwitt ahh
16:17:13 dansmith when we moved flavor info into sysmeta, we caused a giant explosion on every instance list, right around icehouse,
16:17:41 dansmith so we moved to this manual join method where we query the metadata tables separately, but in bulk, to avoid the n:m join
16:18:11 dansmith so the single case is similar, just where n=1, but if m is large enough, still unideal
16:18:16 sean-k-mooney we basicaly need to use a subqurry for the userdata
16:18:29 sean-k-mooney and get the rest of the data normally
16:19:02 dansmith subquery or separate query? not sure how the former will help
16:19:23 sean-k-mooney seperate query
16:19:36 dansmith yeah, we do that on list, just don't on get, that's my point
16:19:48 sean-k-mooney yeah
16:20:12 dansmith all the machinery is there from list, so get can just call it but with the single instance we care about
16:20:18 sean-k-mooney i didnt look a the code specificaly but this sounded familar as i think we already optimised this once before for that list case
16:21:01 sean-k-mooney its just one of those things that works fine for small number and then explode because its n*m complexity
16:21:10 sean-k-mooney in size
16:21:11 dansmith for list, yeah
16:21:31 sean-k-mooney well in this case its joing userdata with each metadata propertiy
16:21:37 dansmith for single get n=1, but m is large
16:21:54 dansmith right
16:22:04 sean-k-mooney ya so its still liniar instad of O(1)
16:22:16 sean-k-mooney not as bad but still kind of sucks
16:22:24 dansmith linearly big is still big :)
16:30:32 openstackgerrit melanie witt proposed openstack/nova stable/ussuri: Follow up for cherry-pick check for merge patch https://review.opendev.org/758463
16:43:25 openstackgerrit Balazs Gibizer proposed openstack/nova master: Ignore PCI devices with 32bit domain https://review.opendev.org/756697
16:45:37 sean-k-mooney gibi: i have seen devices that use teh 32bit domian that can be assigned to qemu
16:45:56 sean-k-mooney althoguh i was using a patched qemu
16:46:31 gibi sean-k-mooney: interesting. Could you check if the patch included support for 32 bit domain? there is a link in the commit message for the qemu code
16:46:54 sean-k-mooney it was an internal intel build for unrelated hardware
16:46:59 sean-k-mooney it was never upstreamed :)
16:46:59 gibi melwitt: improved https://review.opendev.org/756697
16:47:23 sean-k-mooney gibi: looking at your patch i think its fine but it woudl be nice to add some docs for this limitation somewhere
16:47:40 sean-k-mooney maybe in the pci whitelist docs somewhere
16:47:53 gibi sean-k-mooney: we already limit the whitelist to 16 bit
16:47:57 melwitt gibi: thanks, will look
16:48:12 sean-k-mooney gibi: but do we document that limitation or just do it in code
16:49:08 gibi sean-k-mooney: good point, we actually not https://docs.openstack.org/nova/latest/configuration/config.html#pci.passthrough_whitelist
16:49:20 gibi I will add extra information to the config description
16:49:52 sean-k-mooney ya we jsut say "PCI address of the device. Both traditional glob style and regular expression syntax is supported."
16:50:07 sean-k-mooney if we can add a note that we only supprot 16 bit domains ill be happy
16:50:40 gibi sean-k-mooney: I will do
16:51:03 gibi but not today. As it is beer time for me now
16:51:11 gibi see you all tomorrow
16:51:14 gibi o/
16:51:23 sean-k-mooney gibi: redhatters will be off tomorrow
16:51:29 sean-k-mooney its a company day
16:51:34 sean-k-mooney just an fyi
16:52:32 gibi sean-k-mooney: thanks for the heads up, enjoy your free day

Earlier   Later