Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-03-13
15:37:51 artom So TLS if possible, and authentication
15:37:59 BobBall I totally agree.
15:39:08 openstackgerrit Maciej Szankin proposed openstack/nova-specs master: Add use-local-scheduler spec https://review.openstack.org/438936
15:48:51 mriedem dansmith: replied in https://review.openstack.org/#/c/443403/ - let me know if i need to change things. starting orientation for the rest of the day in about 10 minutes
15:50:31 bauzas dansmith: jroll: jaypipes: do we have some spec that explains our consensus in L66 https://etherpad.openstack.org/p/nova-ptg-pike-placement
15:50:32 bauzas ?
15:50:53 bauzas tl;dr: ironic flavors need to be updated by adding extra specs
15:51:12 bauzas s/explains/reflects
15:51:28 jroll bauzas: that's on my list for this week
15:51:36 jroll (writing that spec)
15:51:48 bauzas jroll: okay, ping me when it's done, because I'm trying to push it downstream
15:52:02 jroll bauzas: will do, thanks
15:52:49 bauzas jroll: FWIW, operators would need to add a custom resource class in their extra spec per type of flavor they have, correct?
15:53:22 openstackgerrit Cedric Brandily proposed openstack/nova master: Remove VMware driver _get_vm_ref_from_uuid method https://review.openstack.org/444959
15:54:30 dansmith bauzas: yes
15:54:31 jroll bauzas: yes
15:57:38 openstackgerrit Cedric Brandily proposed openstack/nova master: VMware: use WithRetrieval in get_network_with_the_name https://review.openstack.org/325587
15:59:06 openstackgerrit Cedric Brandily proposed openstack/nova master: VMware: use WithRetrieval in vm_util module https://review.openstack.org/325608
15:59:17 openstackgerrit Cedric Brandily proposed openstack/nova master: VMware: use WithRetrieval in ds_util module https://review.openstack.org/325610
15:59:30 erol sean-k-mooney: Hey, When I added another ironic node, I didt get error from Schedular. I think it is all about RetryFilter.
16:01:16 sean-k-mooney erol: interesting. though the pastbin was showing that the ram filter was the filter that eliminated all of the host
16:01:34 sean-k-mooney erol: Filter results: ['RetryFilter: (start: 4, end: 3)', 'AvailabilityZoneFilter: (start: 3, end: 3)', 'RamFilter: (start: 3, end: 0)']
16:02:35 sean-k-mooney erol: the retry filter eliminated 1 host which was the host that was tried first
16:03:28 erol sean-k-mooney: all of 3 node is kvm node, there is no enough resource that instance need on them. Sorry for the indicate it.
16:04:02 sean-k-mooney erol: so you have a mixed deployment ironic/kvm?
16:04:44 erol I think Retryfilter eliminate Ironic node, and then When it check ram resource it cant find correct one in kvm nodes.
16:04:50 sean-k-mooney and the ram filter failed on the kvm nodes when trying to select a baremetal flavor
16:05:15 erol sean-k-mooney: yeah. 3 Nova Compute 2 ironic compute for now.
16:05:17 sean-k-mooney erol: the retry filter only filters nodes if it tred to scheduler to it and then failed
16:06:18 sean-k-mooney erol: it previously selected u'ironic-compute', u'f01c9155-7d1e-4351-8225-670e0d48452a' that failed to boot and it retryed the rest - that node
16:07:27 erol sean-k-mooney: I am doing lots of mistake about metadata and resources. Then ironic instances faild on provisioning. Maybe i need the delete ironic-compute history when i reach the max attempt.
16:08:09 erol sean-k-mooney: failed*
16:08:11 raghav hi
16:08:37 raghav i am getting error in watcher code..which root causes in nova code
16:09:13 sean-k-mooney erol: perhaps. i have used ironic but only in a limited fashion. one thing that you should be aware of is when deleteing ironic nodes it does not alway delete the neutron port which can cause issues
16:09:49 raghav +----------------+---------------------+-----------------+----------------+-------+ | ID | Hypervisor Hostname | Hypervisor Type | Host IP | State | +----------------+---------------------+-----------------+----------------+-------+ | region!child@1 | praveen-dev | QEMU | 10.223.197.217 | up | +----------------+---------------------+-----------------+----------------+-------+
16:09:50 sean-k-mooney *used ironic much, only in a limited fashion
16:10:01 raghav here id fiels is non-integer
16:10:21 sean-k-mooney raghav: the id field is a string
16:10:35 raghav my nova-api version is 15.0.1
16:10:37 sean-k-mooney raghav: infact its usally a uuid
16:10:57 raghav but in doc its integer
16:11:16 raghav https://github.com/openstack/nova/blob/stable/mitaka/nova/objects/compute_node.py#L59
16:12:18 erol sean-k-mooney: I wish there is a cli command to flush failed attempt for RetryFilter
16:12:23 raghav i am getting parsing error in watcher because of this
16:13:50 sean-k-mooney raghav: so the id field in the db is an int
16:14:20 sean-k-mooney however i belive the ID filed in the api maps to the UUID field in the db as the id filed in the db is not exposed
16:14:51 sean-k-mooney raghav: i could be wrong on that but jaypipes or mriedem shoudl be able to confirm
16:15:26 mriedem which table?
16:15:35 mriedem we expose the compute_nodes.id field out of the API
16:15:37 mriedem b/c we're dumb
16:16:11 mriedem https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/hypervisors.py#L52
16:16:32 mriedem sean-k-mooney: i plan on fixing that
16:16:33 mriedem https://blueprints.launchpad.net/nova/+spec/service-hyper-pci-uuid-in-api
16:17:24 sean-k-mooney raghav: actully i might be getting confused with the flavor sorry ignore what i previously said.
16:17:25 openstackgerrit Mikhail Feoktistov proposed openstack/nova master: Cleanup code in create_ploop_image https://review.openstack.org/444970
16:17:26 sean-k-mooney mriedem: cool
16:17:37 mriedem sean-k-mooney: we expose the flavor.id too out of the API
16:18:55 sean-k-mooney mriedem: yes but for the flavor id that is also user set able as any arbitry string, and i know several people who have reversed the name and id filed on the commandline so that is even more of an issue
16:19:07 openstackgerrit Mikhail Feoktistov proposed openstack/nova master: Cleanup code in create_ploop_image https://review.openstack.org/444970
16:19:26 mriedem sean-k-mooney: well, there is flavor.id and flavor.flavorid
16:19:32 mriedem flavor.flavorid is user-settable
16:19:36 mriedem flavor.id is the primary key
16:20:11 mriedem i guess we don't expose flavor.id here https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/views/flavors.py#L34
16:20:18 mriedem it might be the server get
16:20:49 mriedem https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/views/servers.py#L259 i guess that's flavorid too
16:21:05 sdague mriedem: we don't expose the .id anywhere iirc
16:21:31 mriedem sdague: ok i thought we did and that was part of what prompted your original spec to nest the flavor details in the server response
16:21:46 mriedem it also came up when moving flavors from the cell to the api db
16:21:58 mriedem i might be misremembering details though, it was a year ago
16:23:01 openstackgerrit Artom Lifshitz proposed openstack/nova-specs master: SPICE native client support https://review.openstack.org/442040
16:23:35 sdague mriedem: there were interesting challenges with flavor.id in cells v1, because it is an auto incrementing id
16:23:43 sdague and, thus could be different in the cells
16:24:21 sdague it is used as the join column
16:24:47 dansmith same for cellsv2 of course,
16:24:57 dansmith but I think we've gotten rid of all those things by now
16:28:27 raghav mriedem: so is it bug in nova code?
16:28:47 mriedem raghav: is what a bug in nova code? sorry i'm on a call so i'm not sure what the earlier discussion is about.
16:29:31 mriedem raghav: are you using cellsv1? i see region!child@1
16:29:34 mriedem that looks like cellsv1
16:30:16 raghav mriedem: i have devstack setup
16:30:22 mriedem raghav: note that you don't want to be using cellsv1
16:30:29 raghav mriedem: with ocata release
16:30:35 mriedem raghav: what's your ENABLED_SERVICES variable?
16:30:45 mriedem you shouldn't have n-cell-region or n-cell-child
16:31:32 raghav mriedem: n-api,n-cpu,n-cond,n-sch,n-novnc,n-cauth
16:31:37 mriedem sorry just n-cell
16:36:40 mriedem raghav: in general, we don't want int id's exposed out of the API, which is why we have https://blueprints.launchpad.net/nova/+spec/service-hyper-pci-uuid-in-api
16:36:55 mriedem raghav: but that's going to be done in a microversion
16:37:57 sdague raghav: honestly, I think there is a lot of confusion about how you got to that kind of data, and the only way any of us know that to be possible is if you have a cells v1 setup
16:38:08 sdague raghav: so, perhaps a reproduce scenario would be useful
16:38:15 mriedem raghav: for now os-hypervisors will return an int for the id https://developer.openstack.org/api-ref/compute/?expanded=show-hypervisor-statistics-detail,show-hypervisor-details-detail
16:38:19 mriedem as shown in the docs
16:38:27 mriedem my blueprint is to change that to return the uuid rather than the int id
16:38:48 melwitt sdague: in cells v1, we don't officially support flavors in cell dbs. users like CERN have their own automation that syncs flavors from the api cell db to the child cell dbs. fwiw
16:39:08 mriedem raghav: as for the hypervisor hostname being "region!child@1" that looks like cellsv1
16:39:18 mriedem so it seems you have n-cells running somewhere in your devstack setup
16:39:26 sdague melwitt: right, it was definitely a goofy thing
16:40:25 raghav mriedem: yes its running
16:40:41 raghav mriedem: i could see in screen

Earlier   Later