Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-11-07
13:44:22 sean-k-mooney ok ok so its already been converted and that works cool
13:44:37 sean-k-mooney we had a question about if that would work downstream so i can say yes
13:45:23 mriedem yes, it's also likely how the remaining legacy jobs need to be converted to zuulv3
13:45:37 mriedem since they run the nova/tests/functional/live_migration hook scripts
13:45:58 mriedem and those scripts themselves should probably be broken down into proper ansible tasks, but that's a bigger job
13:46:25 mriedem *gate/live_migration/hooks
13:46:43 sean-k-mooney sure but it give us an easy way to achive the comunity goal and then we can break them donw later
13:46:54 mriedem ^ runs remove ansible commands on the subnode to set it up which we could just do with proper tasks that run on the subnode
13:47:04 mriedem s/remove/remote/
13:49:03 sean-k-mooney when you say set up the subnode you are refering to thing that need to be change post install right
13:49:08 sean-k-mooney e.g. not running devstack
13:50:34 sean-k-mooney if it was runnign devstack that can obviroulsy be done with the native devstack multinode supprot but im guessing they are doing other stuff
13:50:50 mriedem post-install,
13:51:04 mriedem the hook runs some live migration tests post default devstack deploy,
13:51:16 mriedem and then re-configures the controller and subnode for ceph using the devstack-plugin-ceph repo
13:51:26 mriedem that's the part that should probably be converted to proper ansible tasks
13:51:36 sean-k-mooney ah ok
13:52:06 mriedem alternatively we could just ditch that part and have a ceph and non-ceph version of those jobs
13:52:18 mriedem the job used to also configure the nodes for nfs, but then it broke and no one cared to ever fix it
13:52:56 mriedem the pita now is the ceph parts in those jobs aren't working properly when trying to make them run with py3
13:52:59 sean-k-mooney i kind of prefer that option of having two jobs but there is obviously the tradeoof of gate time
13:53:12 mriedem https://review.opendev.org/#/c/692374/
13:53:19 mriedem well, it'd be 4 jobs
13:53:30 mriedem b/c the nova-grenade-multinode and nova-live-migration jobs both use the same scripts
13:54:16 sean-k-mooney right so block migrate and shared storage migthation for both
13:56:00 sean-k-mooney i need to get back to the nvf job too. to test the numa/pinning/hugepages stuff
13:56:57 sean-k-mooney at the moment i have 3 jobs that i need to complete. the nfv job, a ovs-dpdk job and a libvirt/qemu job
13:57:35 sean-k-mooney although we proably wont run them all on each patch to nova
13:59:24 sean-k-mooney mriedem: regarding greade it would be nice if we could use https://review.opendev.org/#/c/548936/
14:17:14 efried sean-k-mooney: gtk
14:39:19 sean-k-mooney im just going for lunch but just wanted to highlight https://review.opendev.org/#/c/683174/4/specs/ussuri/approved/vm-scoped-sriov-numa-affinity.rst
14:39:41 sean-k-mooney that has been up for about a month now with no review.
14:40:20 openstackgerrit Matt Riedemann proposed openstack/nova master: Convert nova-live-migration job to zuulv3 https://review.opendev.org/693364
14:40:49 sean-k-mooney i have it already implemented https://review.opendev.org/#/c/674072/4 and would like to wrap it up before we hit holiday season so if people coudl review that would be awsome
15:12:23 efried sean-k-mooney: would you be offended if I posted a new PS with just spelling corrections before I review?
15:14:50 dansmith efried: sean-k-mooney _can_ actually write english prose if he puts his mind to it
15:15:06 dansmith I've seen it happen
15:15:28 efried I've seen it done, I know it's true. This is closer than usual, but there's no sense putting up a review with a hundred typo comments.
15:15:49 dansmith a hundred typos is closer than usual? ouch :D
15:23:48 mriedem people must be working b/c the gate is load failing a bit again today
15:25:52 openstackgerrit ya.wang proposed openstack/nova-specs master: Re-propose expose-auto-converge-post-copy spec for Ussuri https://review.opendev.org/687199
15:27:41 mriedem lyarwood: the long term solution to crap like https://review.opendev.org/682523 is i think https://review.opendev.org/#/c/650188/
15:27:58 mriedem the imagebackend tells us about shared or not for the root disk storage,
15:28:05 mriedem placement tells us if the local host disk is on shared storage
15:28:17 mriedem no more trying to guess based on imagebackend (wrongly),
15:28:33 mriedem no more trying to create temp files to see if we can on the remote host
15:28:55 mriedem no more shared ssh keys if you're using shared storage,
15:29:10 mriedem and we could schedule within a shared storage pool for things like live migration if you don't use shared ssh keys
15:29:26 mriedem god everybody wins!
15:31:30 lyarwood mriedem: ack yeah agreed, the question about a refactor was actually to mdbooth but yeah ultimatley that spec is going to solve alot of this.
15:35:11 mriedem oh, the *other* matt
15:36:52 lyarwood yup my bad sorry
15:37:14 openstackgerrit Eric Fried proposed openstack/nova-specs master: Add spec for VM-scoped SR-IOV NUMA affinity https://review.opendev.org/683174
15:37:22 efried sean-k-mooney: done ^
15:38:02 openstackgerrit Lee Yarwood proposed openstack/nova master: libvirt: Rename _is_storage_shared_with to _is_path_shared_with https://review.opendev.org/693337
15:39:00 openstackgerrit Merged openstack/nova stable/stein: Revert "openstack server create" to "nova boot" in nova docs https://review.opendev.org/693242
15:41:22 mriedem easy cross-cell-resize change here at the bottom of the stack to keep the series moving https://review.opendev.org/#/c/669012/
15:53:57 efried mriedem: question there, but generally makes sense to me.
15:56:30 sean-k-mooney efried: thanks are you generally ok with it?
15:57:14 efried sean-k-mooney: I didn't actually read it; I'm doing that now. Note that I'm pretty weak on things involving NUMA, SR-IOV, and PCI so I probably won't be able to do more than +1 anyway.
15:58:06 jamesdenton sean-k-mooney I have some hardware here to kick the tires on that, but it may be next week before i have a chance
15:58:08 efried yeah, sorry, when I said "done" I just ment the copy edit
15:58:34 dansmith efried: gave you my answer on mriedem's patch at least (and +2d)
15:58:41 efried ack
15:58:46 sean-k-mooney jamesdenton: cool, 99% is just using code that was already there
15:58:52 dansmith actually had a tab open with a +2 on that from yesterday that I never hit go on
15:59:47 efried dansmith: +A, thanks for confirming.
16:00:07 efried jamesdenton: are you in operator-land? Is sean-k-mooney's use case a thing you care about?
16:02:30 mriedem efried: replied with the big details
16:03:19 jamesdenton i am in operator-land. The use-case doesn't directly impact me, but i try to stay tuned to what's happening with SRIOV for our customers that do use it
16:04:19 mriedem efried: dansmith: thanks
16:04:42 mriedem tell me when my asking for reviews on this series gets too annoying
16:05:34 mriedem also, my wife and daughter just left for disney world (meeting up with my mother in law), you know what that means!
16:05:52 efried mriedem: squeaky wheel and all that. Keep spoonfeeding me the easy ones and I'll keep reviewing them. Still haven't made it all the way through the review guide to hit the heavier ones
16:05:58 dansmith mriedem: eating raw brownie batter shirtless over the sink time!
16:06:07 mriedem close!
16:06:20 mriedem let's just say i still haven't showered or put contacts in yet today
16:06:38 efried I can smell you from here
16:06:39 dansmith get some brownie batter and you've got a party!
16:06:47 mriedem efried: yeah i've been trying to front-load as much of the simpler mechanical and test-only things for that reason
16:06:53 mriedem pants of dance off!
16:06:56 mriedem *off
16:08:14 efried dansmith: I'll take my slice of humble pie from you, but you've got some nerve disagreeing with jacky06 on https://review.opendev.org/#/c/688122/
16:11:14 openstackgerrit Eric Fried proposed openstack/nova-specs master: Add spec for VM-scoped SR-IOV NUMA affinity https://review.opendev.org/683174
16:11:26 efried d'oh, forgot to save my last batch of edits sean-k-mooney ^
16:24:08 dansmith efried: my finger quivered with nerves as I hit the submit button on that one for sure
16:54:36 efried sean-k-mooney: reviewed, +2
16:56:13 sean-k-mooney efried: thanks.
17:07:21 openstackgerrit Matt Riedemann proposed openstack/nova master: Convert nova-live-migration job to zuulv3 https://review.opendev.org/693364
17:45:30 openstackgerrit Eric Fried proposed openstack/nova master: Add emulated TPM support to Nova https://review.opendev.org/631363
17:45:31 openstackgerrit Eric Fried proposed openstack/nova master: WIP: vTPM request_filter https://review.opendev.org/678325
17:45:31 openstackgerrit Eric Fried proposed openstack/nova master: Add support for resize and cold migration of emulated TPM files https://review.opendev.org/639934
18:00:16 sean-k-mooney efried: by the way are you taking over teh vTPM stuff in general this cycle
18:02:06 sean-k-mooney i have added them to my review list in anycase
18:03:02 efried sean-k-mooney: yes. It's actually the whole reason I'm trying to get a devstack with modern libvirt/qemu. It's a nontrivial delta from the previously approved spec as it has to include the encryption bits.
18:03:07 efried and I still don't know wtf that means yet.
18:03:11 efried so, experimenting.
18:03:41 sean-k-mooney oh so the vTPM will actully encrypt its contence on the hsot
18:03:45 sean-k-mooney *host
18:04:04 efried yes, via a virsh "secret"

Earlier   Later