| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-07-07 | |||
| 18:29:26 | mriedem | in pike | |
| 18:29:37 | mriedem | because i need to narrow my focus on what still needs to get done | |
| 18:31:47 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: scheduler uses allocation candidates https://review.openstack.org/476631 | |
| 18:31:48 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: claim resources in scheduler https://review.openstack.org/476632 | |
| 18:32:06 | leakypipes | mriedem, superdan: ^^ | |
| 18:32:43 | leakypipes | definitely should add some functional testing for that. | |
| 18:32:56 | leakypipes | but it's good enough to get reviews on IMO. | |
| 18:34:04 | superdan | did you just kick that out of the gate with that push? | |
| 18:34:09 | superdan | yup | |
| 18:35:42 | mriedem | git review -R | |
| 18:35:44 | mriedem | ftw | |
| 18:35:50 | superdan | yeah | |
| 18:36:34 | leakypipes | mriedem, superdan: yeah, sorry about that :( | |
| 18:37:25 | leakypipes | figleaf had a worry about the previous patch, so perhaps you could see his review on the previous revision of https://review.openstack.org/476632 | |
| 18:39:21 | clarkb | fwiw git review -R shouldn't really be necessary anymore. You should only get a rebase from git review if it hits a merge conflict | |
| 18:39:27 | clarkb | in which case you hvae to push a rebase at some point | |
| 18:42:31 | superdan | clarkb: gerrit identified the bottom as a rebase only. | |
| 18:42:51 | superdan | so I don't think that is accurate for this case | |
| 18:43:12 | leakypipes | superdan: I rebased to master explicitly (by mistake) | |
| 18:43:19 | figleaf | leakypipes: did that comment make any sense? | |
| 18:43:21 | superdan | leakypipes: right I know | |
| 18:43:43 | leakypipes | figleaf: well, I tried to answer as best I could. did you see my response? | |
| 18:43:53 | figleaf | just looking... | |
| 18:44:12 | leakypipes | guys, I need to run out for a couple hours... I will be back later on. sorry! | |
| 18:44:24 | figleaf | leakypipes: np - I have to run out now too | |
| 18:44:45 | figleaf | leakypipes: but I'm still confused as to how this is going to work w/o returning alloc candidates | |
| 19:01:11 | mriedem | superdan: booya fleetify is passing the multinode job http://logs.openstack.org/56/477556/5/check/gate-tempest-dsvm-neutron-multinode-full-ubuntu-xenial-nv/aed9486/ | |
| 19:01:51 | superdan | schwing | |
| 19:02:01 | mriedem | and the migration not found errors are gone | |
| 19:02:32 | mriedem | subnode nova-cpu.conf also has the disable flag http://logs.openstack.org/56/477556/5/check/gate-tempest-dsvm-neutron-multinode-full-ubuntu-xenial-nv/aed9486/logs/subnode-2/etc/nova/nova-cpu.conf.txt.gz | |
| 19:11:39 | mriedem | jgriffith: clarified what i meant in https://review.openstack.org/#/c/330285/88/nova/compute/api.py@3901 | |
| 19:12:24 | openstackgerrit | Sumedh Degaonkar proposed openstack/nova master: Adds 'sata' as a valid disk bus for qemu and kvm hypervisors https://review.openstack.org/459794 | |
| 19:12:43 | jgriffith | mriedem thanks, once they get network back up in the lab I'll go back through and fix the unit tests and pep8 I appear to have broken | |
| 19:12:56 | jgriffith | also will see if I can figure out what broke after the rebase | |
| 19:13:43 | jgriffith | mriedem oh... looks like you didn't see the update | |
| 19:13:56 | jgriffith | mriedem https://review.openstack.org/#/c/330285/89/nova/compute/api.py@3901 | |
| 19:14:03 | mriedem | ha | |
| 19:14:13 | jgriffith | :) | |
| 19:14:19 | mriedem | volume_id doesn't exist | |
| 19:14:24 | mriedem | volume['id'] is what you need | |
| 19:14:39 | jgriffith | awww crap!!! | |
| 19:14:45 | mriedem | otherwise looks good | |
| 19:14:48 | openstackgerrit | Sumedh Degaonkar proposed openstack/nova master: Adds 'sata' as a valid disk bus for qemu and kvm hypervisors https://review.openstack.org/459794 | |
| 19:14:56 | mriedem | let me pad some stats here | |
| 19:15:01 | jgriffith | mriedem thanks, knew I was screwed when I couldn't test | |
| 19:16:50 | mriedem | figleaf: we have some other changes to do in https://blueprints.launchpad.net/nova/+spec/custom-resource-classes-in-flavors right? there is some data migration thing that is supposed to happen on startup of the nova-compute service with ironic, so it can retrofit older flavors | |
| 19:18:59 | mriedem | ah yeah this https://specs.openstack.org/openstack/nova-specs/specs/pike/approved/custom-resource-classes-in-flavors.html#data-model-impact | |
| 19:19:43 | mriedem | superdan: in ^ should that say "will not be schedulable" ? | |
| 19:20:48 | superdan | mriedem: I think it's saying that until you account for the new resource types, | |
| 19:21:13 | superdan | nova will think it could schedule instances to nodes that are already used, because they haven't yet reported that the instance is consuming one of the new resources | |
| 19:21:31 | mriedem | ah | |
| 19:21:45 | superdan | you buy that? | |
| 19:21:54 | mriedem | so CUSTOM_BAREMETAL_INFINIBAND_GOLD is used but not reported as used | |
| 19:22:01 | superdan | yar | |
| 19:22:05 | mriedem | ok, | |
| 19:22:07 | mriedem | so, | |
| 19:22:26 | mriedem | if we move the claim/allocation stuff to the scheduler, can we still do this data migration sync up on nova-compute start? | |
| 19:22:53 | mriedem | i mean, we can, we have the scheduler report client, | |
| 19:23:02 | mriedem | and the update_available_resource periodic, | |
| 19:23:07 | mriedem | just wondering if that will handle it | |
| 19:23:15 | superdan | mriedem: the compute node also isn't going to report any GOLD resources available until it starts up on new code either | |
| 19:23:32 | superdan | so I think if you restart onto the new scheduler, | |
| 19:24:17 | superdan | wait | |
| 19:24:17 | superdan | we'll start asking for resources that expose GOLD once the flavors are updated, but no computes will have any GOLDs to give until they are restarted and start exposing that | |
| 19:24:23 | superdan | did we already do that part in ocata? | |
| 19:24:26 | superdan | yeah, okay yeah | |
| 19:24:31 | superdan | I think I've got this now | |
| 19:24:35 | superdan | lemme try again: | |
| 19:24:57 | superdan | In ocata, we got the ironic driver to expose 1 $ironic_node_resource_class for each node to placement | |
| 19:25:10 | superdan | if we update the real flavors with the new resource type, | |
| 19:25:14 | openstackgerrit | melanie witt proposed openstack/nova master: Make security_group_rules use check_deltas() for quota https://review.openstack.org/477700 | |
| 19:25:16 | openstackgerrit | melanie witt proposed openstack/nova master: Remove 'reserved' count from used limits https://review.openstack.org/446242 | |
| 19:25:17 | openstackgerrit | melanie witt proposed openstack/nova master: Make key_pairs use check_deltas() for quota https://review.openstack.org/477699 | |
| 19:25:20 | openstackgerrit | melanie witt proposed openstack/nova master: Remove useless quota_usage_refresh from nova-manage https://review.openstack.org/446243 | |
| 19:25:21 | openstackgerrit | melanie witt proposed openstack/nova master: Count instances to check quota https://review.openstack.org/416521 | |
| 19:25:25 | superdan | scheduler would send instances to busy computes thinking there was room | |
| 19:25:27 | openstackgerrit | melanie witt proposed openstack/nova master: Make Quotas object favor the API database https://review.openstack.org/410945 | |
| 19:25:28 | superdan | what we need, | |
| 19:25:34 | openstackgerrit | melanie witt proposed openstack/nova master: Add online migration to move quotas to API database https://review.openstack.org/410946 | |
| 19:25:35 | openstackgerrit | melanie witt proposed openstack/nova master: Limit maximum pending resizes for preventing over quota https://review.openstack.org/481739 | |
| 19:25:45 | superdan | is to data migrate the _instance_ flavors to also report that they use one of the new resources, | |
| 19:25:59 | superdan | so that the computes start reporting 1 total, 1 used, 0 available | |
| 19:26:13 | superdan | so really, | |
| 19:26:26 | superdan | ops just need to run data migrations before they update the main flavors to use the override I think | |
| 19:26:31 | superdan | make sense/ | |
| 19:27:31 | mriedem | yeah, that's where i was going in my head, | |
| 19:27:53 | mriedem | that the compute updates the existing instance embedded flavor on startup so we can fix the allocations reported in, | |
| 19:28:02 | superdan | right | |
| 19:28:04 | mriedem | and it's good that none of it turns on until you fiddle the flavors | |
| 19:28:11 | superdan | yeah | |
| 19:28:49 | mriedem | although, we don't actually pull the custom resource off the instance.flavor in the compute, we pull it off the node | |
| 19:29:02 | mriedem | so i'm a bit fuzzy on where the existing instance.flavor comes into play | |
| 19:29:40 | mriedem | this https://github.com/openstack/nova/blob/master/nova/virt/ironic/driver.py#L658 | |
| 19:29:44 | superdan | we report inventory based on the node | |
| 19:29:50 | superdan | we report allocations based on the instance/flavor | |
| 19:30:00 | superdan | and we request from placement based on the real flavor | |
| 19:30:55 | mriedem | ok so RT updates allocations here https://github.com/openstack/nova/blob/master/nova/compute/resource_tracker.py#L1012 | |
| 19:31:52 | mriedem | and this is where we get the allocations off the instance right? https://github.com/openstack/nova/blob/master/nova/scheduler/client/report.py#L141 | |
| 19:32:02 | mriedem | which is only ram, cpu and disk | |
| 19:32:06 | mriedem | not fancy pants ironic stuff | |