Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-07-23
21:10:25 mriedem seems like a glaring omission
21:11:01 mriedem well,
21:11:07 dansmith I dunno,
21:11:10 dansmith we've picked a host/cell
21:11:11 mriedem i guess the idea is that if the user deleted the build request, and we don't create it in a cell,
21:11:13 mriedem then you can't list/show it
21:11:45 dansmith ah right this is the pre-create lookup
21:12:15 dansmith could argue that you should still be able to show deleted I guess
21:12:22 dansmith but not sure it's worth changing now
21:12:46 mriedem heh i guess i added that https://github.com/openstack/nova/commit/8ba92778fe14b47ad4ff5b53022e0550a93f37d3
21:12:51 mriedem i knew it looked familiar
21:12:57 dansmith so on second thought it looks perfect? :)
21:13:05 mriedem it is pretty f'ing nice
21:13:29 dansmith you lost me about two minutes ago since 'Josie' came up on the jukebox
21:13:47 dansmith volume -> 100%, butt -> out of seat
21:14:29 mriedem well it is pretty funky
21:15:18 eandersson What is the general strategy for dealing with hypervisor maintenance (e.g. os updates etc)?
21:15:43 eandersson We wanted to look into a similar implementation to what Amazon does
21:16:12 eandersson Like tagging a hypervisor for downtime
21:16:31 mriedem there used to be a maintenance guide in the ops guide
21:16:32 mnaser i think that's beyond the scope of nova and more like business/operational processes
21:16:42 mriedem but i don't know where the ops guide stuff is being worked on now
21:17:17 mriedem https://github.com/openstack/openstack-manuals/blob/mitaka-eol/doc/ops-guide/source/ops_maintenance.rst
21:17:50 eandersson Yea - I was hoping that maybe someone wrote some exteranl tools for managing this already, or maybe even a feature inbuilt into nova to allow you to set a timer on a compute
21:18:09 mnaser i mean
21:18:10 mriedem didn't we talk about this at some point you and i?
21:18:14 eandersson We did
21:18:16 mriedem you can mark a compute as disabled from scheduling
21:18:26 mnaser and the user will see the status as 'maintenance' or something
21:18:27 mnaser in the api
21:18:27 mriedem https://developer.openstack.org/api-ref/compute/#compute-services-os-services
21:18:37 mnaser i think there is a hoststatus field or something
21:18:41 mnaser unless thats' admin only
21:18:43 mriedem the user doesn't see services
21:18:53 mnaser not but for vms running in that host
21:18:56 mriedem the guests on the host are unaffected if you disable the compute service
21:18:59 mnaser in /servers/<uuid>
21:19:16 mriedem you just can't schedule or migrate new instances to that host while the service is disabled
21:19:57 mriedem you can force the service status to 'down' to evacuate guests from it
21:20:08 mriedem but if it's up and normal otherwise just live migrate
21:20:42 mnaser eandersson: and for customer facing stuff, the nova instance has field 'host_status' which will be 'MAINTENANCE' if the vm is ona disabled host
21:20:49 mnaser so that might be a way to warn consumers that something will happen
21:20:55 mnaser if you wanna say disable it for a few days before
21:21:09 mnaser https://developer.openstack.org/api-ref/compute/#show-server-details
21:21:13 mnaser i swear it's there >:(
21:21:29 mriedem oh right,
21:21:32 mriedem admin only by default
21:21:35 mriedem configurable via policy
21:21:48 mnaser ah yeah, so maybe that might be something you want to mess with policy then
21:21:49 mriedem and only with microversion >= 2.16 which is probably newer than eandersson has
21:22:07 eandersson Yea - we are working hard on upgrading to Queens.. but still stuck on Mitaka for now
21:23:09 eandersson Also, us NUMA pinning everything does not help
21:23:31 mriedem not if you actually want to move things
21:23:50 mnaser while latency goes down with NUMA pinning
21:23:55 mnaser i've noticed overall performance is not as good
21:24:06 mnaser for workloads in general
21:24:21 eandersson Gaming is a bit of a odd one unfortunately
21:25:13 eandersson latency is everything for players (and not only talking about network latency here :D)
21:35:57 mriedem dansmith: i was going to rebase the queued_for_delete series but jaypipes has comments on your migration patch (and the change below it in the series) https://review.openstack.org/#/c/584504/ so you should probably look first
21:43:36 openstackgerrit Dan Smith proposed openstack/nova master: Online data migration for queued_for_delete flag https://review.openstack.org/584504
21:43:49 dansmith (that was just the rebase)
21:48:46 openstackgerrit Dan Smith proposed openstack/nova master: Online data migration for queued_for_delete flag https://review.openstack.org/584504
21:50:34 openstackgerrit Eric Fried proposed openstack/nova master: [placement] Add /reshaper handler for POST https://review.openstack.org/576927
21:50:35 openstackgerrit Eric Fried proposed openstack/nova master: reshaper: Look up provider if not in inventories https://review.openstack.org/585033
21:50:36 openstackgerrit Eric Fried proposed openstack/nova master: Make get_allocations_for_resource_provider sane https://review.openstack.org/584598
21:50:37 openstackgerrit Eric Fried proposed openstack/nova master: Report client: Real get_allocs_for_consumer https://review.openstack.org/584599
21:50:38 openstackgerrit Eric Fried proposed openstack/nova master: Report client: get_allocations_for_provider_tree https://review.openstack.org/584648
21:50:39 openstackgerrit Eric Fried proposed openstack/nova master: Report client: _reshape helper, placement min bump https://review.openstack.org/585034
21:51:45 jaypipes dansmith, mriedem: I'm +2 on em.
21:58:42 mriedem that makes 2 of us
21:59:20 mriedem dansmith: there is an email in the ops ML with some cells v2 questions, i replied with what i could but might be a couple of things in there you want to expand on
22:05:14 openstackgerrit Matt Riedemann proposed openstack/nova master: Update queued-for-delete from the ComputeAPI during deletion/restoration https://review.openstack.org/566813
22:10:23 openstackgerrit Merged openstack/nova master: Add VIFMigrateData object for live migration https://review.openstack.org/515423
22:13:41 mriedem oh wow something merged today
22:16:11 openstackgerrit Matt Riedemann proposed openstack/nova master: doc: link to CERN summit video about upgrading from cells v1 to v2 https://review.openstack.org/585044
22:25:55 openstackgerrit Matt Riedemann proposed openstack/nova master: doc: link to AZ talk from the Rocky summit https://review.openstack.org/585045
23:06:45 openstackgerrit Eric Fried proposed openstack/nova master: WIP: Compute: Handle reshaped provider trees https://review.openstack.org/576236
23:06:46 openstackgerrit Eric Fried proposed openstack/nova master: WIP: Report client: update_from_provider_tree w/reshape https://review.openstack.org/585049
23:48:35 openstackgerrit Matt Riedemann proposed openstack/nova master: Enhance doc to guide user to use nova user https://review.openstack.org/583115
#openstack-nova - 2018-07-25
00:05:09 cfriesen_ how much is "slightly higher"?
00:15:07 eandersson cfriesen_, 2GB for ~128gb and 4GB for 256GB
00:27:43 openstackgerrit Takashi NATSUME proposed openstack/nova master: Remove mox in libvirt/test_driver.py (5) https://review.openstack.org/570842
00:41:23 eandersson cfriesen_, basically in libvirt 1.x virsh nodeinfo and virsh nodememstats shows the same memory size
00:41:34 eandersson but in 2+ the size is different (at least on our machines)
00:43:09 mriedem cfriesen_: shouldn't, unless it's a reschedule
00:43:16 mriedem in which case the compute doesn't pass the reqspec back up
00:43:19 mriedem i have a patch for that
00:43:38 mriedem cfriesen_: https://review.openstack.org/#/c/582417/
00:43:41 mriedem needs an rpc api change
00:45:06 mriedem cfriesen_: but starting in queens we wouldn't call back to the scheduler on a reschedule because we have alternate hosts from the top level scheduling call
00:45:20 mriedem and any instance created after newton would have a request spec, so that first RequestSpecNotFound block shouldn't be hit
00:46:39 mriedem cfriesen_: yeah i can't see how http://paste.openstack.org/show/726559/ is needed b/c the request spec has the original pci requests and that's what the scheduler filter uses
00:46:52 mriedem and anything on the compute which claims using instance.pci_requests would lazy-load it
01:11:03 Kevin_Zheng mriedem, do you have a minute? I need some help on https://review.openstack.org/#/c/583927/
01:26:00 openstackgerrit Takashi NATSUME proposed openstack/nova master: Remove mox in libvirt/test_driver.py (6) https://review.openstack.org/571330
01:26:55 openstackgerrit Takashi NATSUME proposed openstack/nova master: Remove mox in libvirt/test_driver.py (7) https://review.openstack.org/571992
01:27:05 openstackgerrit Takashi NATSUME proposed openstack/nova master: Remove mox in libvirt/test_driver.py (8) https://review.openstack.org/571993
01:27:36 openstackgerrit Merged openstack/nova master: libvirt: use dest host vif migrate details for live migration https://review.openstack.org/551370
02:44:31 openstackgerrit Takashi NATSUME proposed openstack/nova master: Transform volume.usage notification https://review.openstack.org/580345
02:44:48 openstackgerrit Takashi NATSUME proposed openstack/nova master: Transform libvirt.error notification https://review.openstack.org/484851

Earlier   Later