Earlier  
Posted Nick Remark
#openstack-nova - 2017-10-18
13:29:54 cdent efried: going back through the log looking at the conversation you had with sean-k-mooney and rgerganov; where in the powervm universe is the mapping from rp uuid to <other> going to live?
13:30:42 stephenfin efried: The problem I'm trying to get at is, even if we somehow managed to keep non-PCI-needing instances off of PCI-having NUMA nodes, we can still end up in the situation where we have no free PCI-having NUMA nodes for PCI-needing instances
13:31:09 mdbooth lyarwood: Could it be cause there's no request context at that point?
13:31:18 mdbooth i.e. no user, no api call?
13:31:26 efried cdent That's a pretty big TBD. We don't have any persistent data on our "hypervisor" (the NovaLink partition), which is a fairly fundamental point of architecture. You're supposed to be able to trash the partition and recreate it with no loss.
13:31:32 stephenfin This approach reduces the possibility but it doesn't mitigate it entirely. We want to use the two in tandem, which is what you've kind of hinted at (I think)
13:31:45 lyarwood mdbooth: yeah sorry, https://github.com/openstack/nova/blob/master/nova/context.py#L279
13:31:59 sean-k-mooney stephenfin: that was the usecase we created https://github.com/openstack/nfv-filters/blob/master/nfv_filters/nova/scheduler/filters/aggregate_instance_type_filter.py to address
13:32:11 openstackgerrit Chris Dent proposed openstack/nova-specs master: Add spec for symmetric GET and PUT of allocations https://review.openstack.org/508164
13:32:16 efried stephenfin Yes. You'll never get around that limitation. Unless you want to totally deny non-PCI-needing instances from booting on PCI-having NUMA nodes. Which ain't reasonable IMO.
13:32:43 sean-k-mooney it allows you to create aggregates of nodes with scarce resouces and require that they are requested in the flavor to schedule to those nodes
13:32:55 stephenfin efried: But you will with this spec, which allows booting PCI-needing instances from using non-PCI-having NUMA nodes
13:32:58 efried cdent Off the cuff, the idea would be to make some part of the RP relate to some part of whatever thingy we're "mapping" to.
13:33:16 sean-k-mooney this filter will work with traits in request in the flavor too by the way
13:33:19 mdbooth lyarwood: Yeah, that would be required for a glance request, I guess
13:33:24 stephenfin sean-k-mooney: I saw that, but it wasn't too inflexible, hence https://blueprints.launchpad.net/nova/+spec/reserve-numa-with-pci
13:33:41 mdbooth Although there's useful work it can do without that
13:33:50 sean-k-mooney that is your new weigher right
13:34:00 lyarwood mdbooth: it's the cinder encryption metadata lookup that's failing at the moment
13:34:08 sean-k-mooney stephenfin: i tought that merged in pike?
13:34:24 efried stephenfin The "problem" you won't get around is where I boot a hundred non-PCI-needing instances and run out of non-PCI-having NUMA nodes after the first fifty, so I start using my PCI-having NUMA nodes. Once those are full, I can no longer boot "required"-affinity PCI-needing instances.
13:34:28 stephenfin sean-k-mooney: Correct. If you had to divide your cloud into aggregates for PCI devices, it made it very inflexible if you wanted to, say, start using a lot of PCI-having instances or vice versa
13:34:43 stephenfin sean-k-mooney: It did. I'm saying https://review.openstack.org/#/c/361140/ forms a one-two combo :)
13:35:03 mdbooth lyarwood: Hmm, yeah. That's not gonna work unless we've got some special-sauce cinder admin login, or we cached it.
13:35:57 stephenfin efried: Yup, there's nothing we can do about that, I'm afraid. If you want to use 'required' then this is what you signed up for. Go buy more PCI devices :P
13:36:03 sean-k-mooney stephenfin: right we really need to merge https://review.openstack.org/#/c/361140/29/specs/queens/approved/share-pci-between-numa-nodes.rst and implemented this cycle it was originally ment to work in icehouse..
13:36:33 efried stephenfin Cool, then we're on the same page mentally; but then I don't think that chunk of the spec is expressing any reasonable "Alternatives".
13:36:44 cdent efried: k, thanks. I’d rather we avoid putting that kind of mapping into placement if possible, but it seems there is a fairly generic need for a translation. I presumem using rp.name is insufficient. Next comes the sort of config info that rgerganov mentioned (things like the IP for the shared storage piece?). That, to me, especially should not go into placement as we’re doing a serious mission creep at that point.
13:36:46 stephenfin efried: However, the https://blueprints.launchpad.net/nova/+spec/reserve-numa-with-pci should help mitigate that, and the 'preferred' policy in https://review.openstack.org/#/c/361140/ should solve it completely
13:36:54 sean-k-mooney stephenfin: as an employee of a hardware vendor that sells pic devices i feel oblidged to +1 that responce :P
13:37:02 stephenfin *if* you're ok living with the downsides (possible performance impact)
13:37:13 stephenfin sean-k-mooney: :P
13:37:31 stephenfin efried: Sure, I'm happy to drop it, if that makes sense? :)
13:37:46 stephenfin *if that would help?
13:37:57 efried stephenfin If you don't mind, perhaps give me a crack at rewording it?
13:38:04 stephenfin efried: Knock yourself out
13:38:32 openstackgerrit Matt Riedemann proposed openstack/nova-specs master: Add pagination and timestamp filtering support for os-migrations API https://review.openstack.org/506030
13:38:46 markus_z Here's a doc change waiting for review for 5 weeks: https://review.openstack.org/#/c/476188/
13:38:55 stephenfin efried: Also, saw your reply on the "two spaces indicate age" comment. It was just an amusing article I read once upon a time that stuck with me. Not an actual ageist thing or anything!
13:39:19 efried cdent I agree with you (at least in the sense that I've accepted the stated limitations of the placement architecture). We could conceivably get some mileage out of overloading the 'name' field, or embedding information into gross custom traits. But I would be wanting to look for cleaner alternatives pretty hard before resorting to that.
13:39:21 stephenfin markus_z: Sure, I'll take a look
13:39:27 markus_z stephenfin: thanks!
13:39:58 openstackgerrit Stephen Finucane proposed openstack/nova master: docs: Explain the flow of the "serial console" feature https://review.openstack.org/476188
13:41:33 efried stephenfin No offense taken, of course. It's an argument that's been made to me before, so you may have caught some blowback from arguments I've had with others who were... um... more serious about it (to the point of -1ing stuff). My take on it is that the reasons for "never do that" don't apply to programmers.
13:42:55 dansmith mriedem: tell me you don't care and I'll +W: https://review.openstack.org/#/c/512406/3
13:44:17 openstackgerrit Ghanshyam Mann proposed openstack/nova-specs master: Spec for API extensions policy removal https://review.openstack.org/508101
13:44:48 efried stephenfin Before I start, can you confirm that https://review.openstack.org/#/c/361140/29/specs/queens/approved/share-pci-between-numa-nodes.rst@131 was indeed supposed to be "without"?
13:45:29 gmann johnthetubaguy: sdague alex_xu ^^ updated spec - https://review.openstack.org/#/c/508101/
13:45:46 mriedem dansmith: i could fix quick, but don't care too much
13:45:48 gmann janki: where i can see the nova logs? i see all networking things there
13:46:10 stephenfin efried: Ah, cool, just making sure. I don't do it because it was never something that was taught to do, though I have been known to strip them out when I rewrite stuff. All good point though
13:46:15 stephenfin efried: Yes, without
13:46:18 dansmith mriedem: if it were master I'd want it fixed, but I guess you could make the argument that it doesn't matter on a frozen branch
13:46:38 peter-hamilton hi everyone, i'm hoping to get final feedback on the updated cert validation spec: https://review.openstack.org/#/c/488541/
13:46:44 mriedem dansmith: i'll fix it quick, 2 minutes
13:46:45 peter-hamilton let me know if you have any questions
13:46:48 janki gmann, https://logs.opendaylight.org/releng/jenkins092/netvirt-csit-1node-openstack-pike-upstream-stateful-carbon/36/control_1/n-api.log.2017-10-16-060444.gz
13:47:15 mdbooth dansmith: The problem with that is the hair pulling when some random other test fails because of it, and then you end up having to bisect a testrun to determine the ordering which causes a failure.
13:47:21 janki gmann, nova-cpu logs https://logs.opendaylight.org/releng/jenkins092/netvirt-csit-1node-openstack-pike-upstream-stateful-carbon/36/compute_1/n-cpu.log.2017-10-16-060456.gz
13:47:30 dansmith mdbooth: that's what I said in my review, yes
13:47:51 dansmith mdbooth: the thing is that it's in a branch that shouldn't really get a lot of debug anymore
13:48:04 mdbooth dansmith: I was agreeing with you, but with additional angst.
13:48:35 mdbooth lol
13:48:56 johnthetubaguy gmann: there is a bit of wording in there I am not totally sure about, agreed with what I think you mean.
13:50:06 gmann johnthetubaguy: right, ll update thanks
13:50:18 gmann janki: error is floating ip pool is not found
13:51:00 janki gmann, yes. because the API call to create it fails right.
13:51:21 gmann janki: API fail because there is no floating ip pool in your env
13:52:06 janki gmann, isnt that what /compute/v2.1/os-floating-ips do?
13:52:36 gmann janki: you mean POST?
13:53:00 janki gmann, ya. POST on compute/v2.1/os-floating-ips returns 404
13:53:01 gmann janki: error is raised from here but i can double check the logs to confirm them same - https://github.com/openstack/nova/blob/fce56ce8c04b20174cd89dfbc2c06f0068324b55/nova/network/neutronv2/api.py#L1985
13:54:37 janki gmann, https://logs.opendaylight.org/releng/jenkins092/netvirt-csit-1node-openstack-pike-upstream-stateful-carbon/36/tempest/tempest.log.gz - search for os-floating-ips
13:56:23 openstackgerrit Matthew Booth proposed openstack/nova master: libvirt: Don't VIR_MIGRATE_NON_SHARED_INC without migrate_disks https://review.openstack.org/507202
13:58:03 dansmith mriedem: fyi, this is the last thing i think is critical to land for fixing up our weird cell0 listing wart: https://review.openstack.org/#/c/508595/8
13:58:14 dansmith the smartness patches after that are not critical, just gravy
13:58:38 dansmith that jenkins -1 isn't going to disappear, in case that has been deterring review
13:58:50 openstackgerrit Matt Riedemann proposed openstack/nova stable/ocata: libvirt: add check for VIR_DOMAIN_BLOCK_REBASE_COPY_DEV https://review.openstack.org/512406
13:58:51 mriedem dansmith: ^
13:59:07 mriedem dansmith: ok, what's been deterring review is the looming newton eol and spec freeze
13:59:21 dansmith mriedem: ack
13:59:47 gmann janki: yea floating ip pool is needed to create the floating ip and if there is nothing then raise error.
14:00:07 gmann janki: what is value of default_floating_pool in conf? under default or neutron section
14:00:28 openstackgerrit Balazs Gibizer proposed openstack/nova master: Extract instance allocation removal code https://review.openstack.org/513041
14:00:34 dansmith mriedem: then speaking of spec freeze: https://review.openstack.org/#/c/510244/
14:00:39 openstackgerrit Merged openstack/nova-specs master: Add pagination and timestamp filtering support for os-migrations API https://review.openstack.org/506030
14:00:46 dansmith we *have* to have that for a variety of things
14:01:16 dansmith mriedem: it's a thick review, so you might just want to assume that efried has worded the ass off it and stamp it through
14:02:03 efried stephenfin Are you still looking at ^?
14:03:10 janki gmann, its "public" under [DEFAULT] In nova.conf
14:03:28 janki gmann, these will also depend on values in tempest.conf right?
14:03:34 stephenfin efried: At what now?
14:03:46 efried stephenfin The granular resource request spec
14:04:14 stephenfin efried: It's on my backlog, but I think I saw it merge this morning?
14:04:23 stephenfin Or at least get some +2s?
14:04:51 efried stephenfin Not merged yet, has dansmith +2 and some +1s.
14:05:42 gmann janki: hat tests did not pass the pool so default is being used and not found in neutron
14:05:55 gmann *that
14:07:15 janki gmann, ohhkk..so what next? I still doubt if tempest.conf has anything to do with this
14:08:17 stephenfin efried: Then yes, I should get to it before EOD

Earlier   Later