Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-10-17
15:31:26 bauzas my main concern is that a modification would require two places
15:32:25 mriedem the only duplication is L698-L706 https://review.openstack.org/#/c/510938/5/nova/compute/manager.py
15:32:35 bauzas I agree
15:32:51 mriedem whch is also in _delete_allocation_after_move now
15:32:51 mriedem :)
15:33:03 mriedem and _rollback_live_migration
15:33:17 bauzas mriedem: but what if someone wants to add another call ?
15:33:33 mriedem so how about we do the consolidation into a util in a follow up?
15:33:47 bauzas I'm fine with that
15:33:50 mriedem hell, we could just do a private method inthe compute manager itself
15:33:52 bauzas I haven't reviewed the change
15:33:53 mriedem if you don't want to touch the RT
15:34:03 bauzas I just wanted to open a discussion
15:34:26 mriedem i'm cool with refactoring out the duplication, but not in this change since it has to be backported
15:34:27 bauzas but yeah, if we were doing the consolidation, would be worth doing all of it in a single patch
15:34:35 bauzas that's a good point
15:34:47 bauzas ok, I'll leave a comment and +W
15:45:02 mriedem bauzas: if you've got the time, we also need these stable/pike backports https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:stable/pike+topic:bug/1715182
15:45:15 mriedem lyarwood: ^
15:45:27 bauzas mriedem: lyarwood is off sick
15:45:38 mriedem too much volume encryption will do that
15:47:15 bauzas I have an internal meeting until 4pm UTC, so I guess I have 13 mins of free time now :)
16:00:04 melwitt mriedem: I'm okay with this libvirt driver change now, to detach and reattach volumes during reboot. apparently there have been summit and PTG discussions about doing that in hard reboot for some time, so if that's the plan anyway, it seems fine to me https://review.openstack.org/#/c/400384
16:00:36 mriedem i don't remember those discussions
16:02:36 dansmith I do, it was around the need to re-fresh the connection info for monitor ip changes, IIRC
16:02:54 mriedem that's not this though
16:04:16 dansmith I haven't looked at the patch, but okay
16:04:31 dansmith we did talk about re-building the storage connections on reboot for that reason though
16:05:21 mriedem yeah i remember that, and have the todo for it still
16:08:09 mriedem ok so we have an encrypted volume test in tempest, i could just modify that to hard reboot the instance after an ecrypted volume is attached and it should blow up the instance right?
16:08:10 dansmith efried: did alex_xu get everything resolved that he was concerned about on that spec? I saw a bunch of discussion here
16:08:40 melwitt mriedem: I think so, yes
16:09:02 mriedem melwitt: ok in that case i'm going to tweak tempest to recreate the failure and then make it depend on the nova fix
16:09:30 mriedem melwitt: jbernard: https://review.openstack.org/#/c/512622/ is passing
16:09:38 mriedem we need to get that in asap to unblock nova stable
16:09:43 mriedem jdurgin: ^
16:09:59 melwitt mriedem: sounds like a good idea
16:10:25 melwitt mriedem: you did a Depends-On to see it pass on stable?
16:10:59 mriedem melwitt: i can quick
16:11:25 melwitt okay, just wanted to make sure that's going to work for you
16:11:47 efried dansmith Not really. We need another writeup. OTP, give me a few mins.
16:11:53 openstackgerrit Matt Riedemann proposed openstack/nova stable/ocata: libvirt: add check for VIR_DOMAIN_BLOCK_REBASE_COPY_DEV https://review.openstack.org/512406
16:11:55 mriedem melwitt: ^
16:12:10 dansmith efried: more in this spec? i thought he was suggesting another spec.. but sure, np
16:13:56 efried dansmith Which spec are we talking about?
16:14:19 dansmith efried: your granular request word-a-palooza
16:16:24 efried dansmith Oh, no, that should be solid as is. Whatever else needs to come out of that will have to go... somewhere else.
16:16:45 dansmith efried: okay that's what I thought, but wanted to clarify
16:17:00 dansmith so I was going to drop my +2 on there, but didn't want to if alex_xu had more things implied from his +0 earlier
16:33:20 jbernard mriedem, melwitt: +A'd, thanks ya'll
16:33:50 melwitt jbernard: awesome, thanks
16:33:55 mriedem sean-k-mooney: comments inline https://review.openstack.org/#/c/502306/
16:34:15 mriedem sean-k-mooney: honestly i think that's going to be a very tough mountain to climb at this point given all of the dependencies
16:34:32 mriedem sean-k-mooney: there seems to be a fundamental mistaken assertion in there that nova creates ports before a host is chosen
16:34:34 mriedem which is not the case
16:34:45 mriedem jbernard: thanks
16:35:12 sean-k-mooney mriedem: i taught nova called the create_or_get port api before calling placement?
16:35:40 sean-k-mooney mriedem: but thank for reviewing. yes this does have a lot of dependncies
16:35:40 mriedem sean-k-mooney: no, ports are created during allocate_for_instance in the compute service
16:35:44 mriedem after a host is chosen by the scheduler
16:36:13 mriedem johnthetubaguy has an older spec to move port creation to conductor, which might be another dependency for this
16:36:33 mriedem sean-k-mooney: isn't the qos policy applied to the port?
16:36:34 sean-k-mooney mriedem: but how will it know if it should be creating a sriov port or normal port if it only did it at the compute node
16:36:44 mriedem sean-k-mooney: nova doesn't create sriov ports
16:36:50 mriedem nova creates regular old boring ports
16:37:00 mriedem if you need a fancy nfv port, you create it in neutron and provide it when creating the server
16:37:28 mriedem if the qos policy is applied to the port, and nova creates the port, then i think we assert that nova-created ports don't have minimum bandwidth guarantees
16:37:30 mriedem easy peasy
16:37:32 sean-k-mooney mriedem: yes the qos policy is on the port but it can also be on the network which adds it to the port automatically
16:37:40 mriedem if you want the fancy, you precreate the port
16:37:53 mriedem ok then that gets complicated,
16:38:03 sean-k-mooney mriedem: yes you are right sriov or qos implies you precreate the port
16:38:05 mriedem because you can create a server with a nic uuid and nova will create the port in that network
16:38:21 johnthetubaguy that sounds a lot like the routed network case that triggered the move to conductor thingy
16:38:38 mriedem and if the qos policy on the network doesn't match the compute host i guess you have problems
16:38:41 mriedem johnthetubaguy: yes
16:40:15 johnthetubaguy sean-k-mooney: if you want to take on the refactor it is now I think just a case of moving the code, we create missing ports (the conductor bit), then update all the ports (the compute bit)
16:40:16 mriedem the spec also doesn't really mention any details about move operations, which we know are the hairiest thing to deal with
16:41:06 mriedem because during a move the allocations have to be made on the target host and eventually removed from the source host - is the neutron agent going to screwed up by that?
16:41:12 mriedem *to get
16:41:59 sean-k-mooney mriedem: it should not as neutron will not be doing allocations from placement iteself
16:42:30 sean-k-mooney mriedem: but that said we have not gone into details of the move operations in the spec
16:42:52 melwitt stephenfin: I refactored the unit tests to address your comment https://review.openstack.org/#/c/498983
16:43:36 mriedem sean-k-mooney: that was another thing, the spec loosely talked about modifying the nova resource tracker, i think to create allocations for these bandwidth things, but we want the scheduler to create allocations, not the compute (RT)
16:44:09 mriedem because when you move the instance, the source host RT might overwrite the allocations for the instance which are meant for the dest host
16:44:27 mriedem because of periodic task race window wackiness
16:45:25 sean-k-mooney mriedem: yes in the future we defiently want to move the allocation out of the compute node the sepc was assuming the workflow we have today where the compute node does the allocation but i dont think it would break if the allocation was done earlier
16:46:19 mriedem sean-k-mooney: if today is >= pike, we don't create allocatoins in the computes anymore
16:46:31 mriedem allocations in the computes is soooooo 2 months ago :)
16:46:38 sean-k-mooney mriedem: oh did that land in pike cool
16:46:47 mriedem yeah i linked to the change in the spec
16:47:05 mriedem as a result of that change we've been doing a lot of cleanup of places where we dont cleanup allocations
16:47:10 sean-k-mooney mriedem: so is that now done in the conductor/scheduler
16:47:10 mriedem still working on getting those fixes into stable/pike
16:47:18 mriedem the scheduler claims yes
16:47:27 mriedem for VCPUS/MEMORY_MB/DISK_GB
16:47:43 mriedem if you have NUMATopologyLimits, those get claimed in the compute still,
16:47:53 mriedem since we don't model NUMA via nested RPs in placement yet
16:48:18 sean-k-mooney pci device im guessing are still on compute node too for same reason
16:48:40 mriedem yeah

Earlier   Later