Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-07-01
14:32:52 openstackgerrit Martin Midolesov proposed openstack/nova master: Implementing graceful shutdown. https://review.opendev.org/666245
14:49:01 openstackgerrit Balazs Gibizer proposed openstack/nova master: nova-manage: heal port allocations https://review.opendev.org/637955
14:49:01 openstackgerrit Balazs Gibizer proposed openstack/nova master: WIP: Add rollback to heal port allocation https://review.opendev.org/668184
15:02:47 openstackgerrit Matt Riedemann proposed openstack/nova stable/rocky: Disable limit if affinity(anti)/same(different)host is requested https://review.opendev.org/659246
15:08:17 kashyap mriedem: Thanks! I'm blinder than a mole.
15:08:49 kashyap Guess it's one of those fully useful timeouts.
15:27:30 mriedem lyarwood: can you take a look at these queens backports? https://review.opendev.org/#/q/status:open+project:openstack/nova+branch:stable/queens+topic:bug/1830747
15:32:38 openstackgerrit Merged openstack/python-novaclient master: Blacklist sphinx 2.1.0 (autodoc bug) https://review.opendev.org/668398
15:35:31 mriedem need another non-lee stable core for this queens backport https://review.opendev.org/#/c/662471/
15:50:52 openstackgerrit Balazs Gibizer proposed openstack/nova master: WIP: Add rollback to heal port allocation https://review.opendev.org/668184
15:51:45 gibi mriedem: what do you prefer? Should I squash the rollback code ^^ into the heal port allocation patch or keep it separate
15:54:51 mriedem gibi: we can drop the code in the original change that generates the list of manual commands to run if we have the rollback stuff right?
15:55:19 gibi mriedem: no, we have to tell the admin what to do when whe rollback fails
15:55:20 mriedem i'd prefer that we rollback and drop the manual list of commands that get generated if possible
15:55:55 gibi mriedem: I mean I can drop it from the original patch and re-introduce it in the rollback patch
15:56:01 gibi if you prefer
15:58:37 mriedem to be clear, the only manual set of commands you plan on dumping are for the port update rollbacks, right? not the thing for the allocations.
15:59:56 mriedem because now we'll be doing: 1. port updates - if good, put allocations, else rollback port updates, 2. if put allocations is good, we're done for that instance, else rollback port updates
15:59:56 gibi commands are only for cleaning the allocation key from the binding:profile
16:00:42 efried I was thinking, does it cost us anything to *gather* those CLIs? Would be useful to print them if you selected --don't-heal-port-allocations-I'll-do-it-myself
16:00:44 mriedem right so we don't need https://review.opendev.org/#/c/637955/29/nova/cmd/manage.py@1876 and _port_update_to_openstack_cli if we move the port updates to before the PUT allocatoins call
16:01:19 mriedem efried: you mean --skip-port-allocations ?
16:01:23 mriedem we don't even hit this code if you specify that
16:01:24 efried yeah
16:01:27 efried i know
16:01:29 efried swhat I'm saying
16:01:30 mriedem and i really don't want to chase CLIs
16:01:52 gibi mriedem: we need a very similar piece of code in the rollback code https://review.opendev.org/#/c/668184/3/nova/cmd/manage.py@1947
16:02:54 gibi mriedem: so I can drop https://review.opendev.org/#/c/637955/29/nova/cmd/manage.py@1876 from that patch, but then I will re-introduce a similar CLI generating code in the rollback patch https://review.opendev.org/#/c/66818
16:03:14 gibi mriedem: or shall I only tell the admin to clean the allocation key but not generate the CLI for it?
16:04:50 gibi efried: if we go and try to heal the port allocations then generating CLIs is cheap. But if we are asked not to heal port allocations with --skip-port-allocations then we don't even call neutron to see if there is anything to heal. So in that case it is expensive to generate CLIs
16:05:11 mriedem i just don't really like generating CLIs from a CLI
16:05:40 mriedem i think i'd prefer to describe something in https://docs.openstack.org/nova/latest/admin/support-compute.html and just link to that
16:06:40 mriedem i think i'd also prefer to squash the changes
16:06:45 gibi mriedem: I can dump the port id's that needs to be updated and link to https://docs.openstack.org/nova/latest/admin/support-compute.html where I write a CLI template. But I don't see why I the admin needs to go through these hops to run a CLI at then end
16:07:35 mriedem it seems easier and more future proof to document the situation and recovery solution in the docs than in code is all
16:07:48 mriedem if this needs to change, it would be easier to do it in the docs i think
16:08:04 gibi mriedem: OK, I will go with your suggestion and write the CLI in the doc. (and squash the changes)
16:08:07 gibi mriedem: thanks
16:10:11 mriedem stephenfin: a couple of comments in this queens backport https://review.opendev.org/#/c/629597/
16:15:40 efried gibi: You going to fold in https://review.opendev.org/#/c/668062/ as well?
16:16:30 mriedem just rebase on top of it
16:16:33 mriedem don't squash it in
16:17:29 openstackgerrit Matt Riedemann proposed openstack/nova stable/rocky: Fix double word hacking test https://review.opendev.org/665788
16:25:13 openstackgerrit Merged openstack/nova stable/stein: Delete resource providers for all nodes when deleting compute service https://review.opendev.org/666840
16:25:20 openstackgerrit Merged openstack/nova stable/stein: Init HostState.failed_builds https://review.opendev.org/668282
16:27:00 openstackgerrit Matt Riedemann proposed openstack/nova stable/rocky: Init HostState.failed_builds https://review.opendev.org/668520
16:29:13 openstackgerrit Matt Riedemann proposed openstack/nova stable/queens: Move get_pci_mapping_for_migration to MigrationContext https://review.opendev.org/661571
16:33:02 openstackgerrit Merged openstack/nova stable/rocky: Grab fresh power state info from the driver https://review.opendev.org/667955
16:37:32 openstackgerrit Matt Riedemann proposed openstack/nova stable/queens: Allow driver to properly unplug VIFs on destination on confirm resize https://review.opendev.org/661572
16:41:47 mriedem lyarwood: you can pull the trigger on https://review.opendev.org/#/c/661572/
16:41:54 mriedem and then that's everything for queens and the next release
16:42:10 mriedem https://review.opendev.org/#/q/status:open+project:openstack/nova+branch:stable/queens+label:Code-Review=2
16:42:34 mriedem melwitt: can you review this queens backport? https://review.opendev.org/#/c/662471/
17:29:40 admin0 hi all .. in horizon, right after login, in the usage summary, when I enter the date and click submit, the data comes out .. how is that data taken out? is it some sort of nova/compute api ?
17:57:48 sean-k-mooney admin0: i think its from the simple tenant usage api endpoint
17:58:24 sean-k-mooney admin0: https://developer.openstack.org/api-ref/compute/?expanded=show-usage-statistics-for-tenant-detail#show-usage-statistics-for-tenant
18:01:19 admin0 sean-k-mooney, thanks
18:13:44 sean-k-mooney mriedem: by the way i am bumping the minium version fo os-traits in https://review.opendev.org/#/c/666914/ too. should we split that out into its own change so we both can depend on it too avoid our prefilter changes conflicting?
18:14:11 melwitt mriedem: can do
18:18:28 gibi efried, mriedem: I will rebase on top of https://review.opendev.org/#/c/668062/
18:21:59 sean-k-mooney efried: for https://review.opendev.org/#/c/651269/4/specs/train/approved/libvirt-pmu-configuration.rst@11 i understand what you ment by topic and commit message but what did you mean by filename? it already matches correct
18:22:04 sean-k-mooney fixing the other two now
18:22:31 efried sean-k-mooney: yup
18:23:53 sean-k-mooney oh you are just saying make all 3 match
18:24:31 openstackgerrit sean mooney proposed openstack/nova-specs master: Libvirt: add vPMU spec for train https://review.opendev.org/651269
18:25:04 sean-k-mooney ok they should all match now
18:26:09 mriedem sean-k-mooney: i'm not too worried about conflicts at this point
18:26:58 sean-k-mooney ok was just wondering if i made sense to have the requiremetns bump be its own thing but its also trival to fix anyway so ill leave it for now
18:29:45 Nick_A does kvm instance I/O get "paused" so to speak during snapshots by default?
18:30:10 sean-k-mooney it depends on the image backend
18:30:26 sean-k-mooney but generally yes
18:30:30 sean-k-mooney at least file io is
18:31:27 sean-k-mooney we quiesce disk io the guest while we do the online snapshot
18:32:01 sean-k-mooney network io should not be affected unless it would have to write to disk as part of the packet processing
18:32:28 sean-k-mooney i think we rely on the qemu guest agent to do this
18:32:49 sean-k-mooney at least when using local qcow2 images
18:33:21 sean-k-mooney for rbd or cinder backed instances the storage backend does the snapshot and im not sure if we quiesce the guest or not
18:33:37 mriedem efried: guess what https://review.opendev.org/#/c/641138/1//COMMIT_MSG@16
18:33:48 mriedem https://review.opendev.org/#/c/666959/1/nova/tests/functional/regressions/test_bug_1825537.py@51
18:34:32 Nick_A thanks sean-k-mooney
18:34:51 efried mriedem: waitwait, do I get an I-told-you-so??
18:34:55 mriedem you do
18:35:01 efried f'in aye
18:35:06 efried okay, I'm done.
18:35:20 mriedem luxuriate
18:36:51 Nick_A could you also explain why snapshots take up twice the instance's used disk space on compute before they're uploaded to glance?
18:37:16 lyarwood mriedem: sorry was out, just getting to the stable reviews you pinged about earlier
18:39:14 Nick_A by that I mean if there's an instance using 4.9G, the snapshots dir will have two files of 4.9G each before upload
18:41:10 sean-k-mooney Nick_A: one is likely the snapshot and the other the root disk
18:41:33 sean-k-mooney i have not really looked at the code so not sure
18:42:46 Nick_A Why would they be the same size then?
18:43:24 mriedem Nick_A: the source code is freely available so you could also dig into this
18:43:45 mriedem this isn't really a support channel
19:19:48 openstackgerrit Matt Riedemann proposed openstack/nova stable/stein: Add functional recreate test for regression bug 1825537 https://review.opendev.org/666959
19:19:48 openstack bug 1825537 in OpenStack Compute (nova) stein "finish_resize failures incorrectly revert allocations" [Medium,In progress] https://launchpad.net/bugs/1825537 - Assigned to Matt Riedemann (mriedem)
19:19:48 openstackgerrit Matt Riedemann proposed openstack/nova stable/stein: Drop source node allocations if finish_resize fails https://review.opendev.org/667155
19:29:28 mriedem sean-k-mooney: left an idea in https://review.opendev.org/#/c/667177/ on how to avoid the driver interface change for backports
19:29:48 mriedem sean-k-mooney: heh you thought of the same, "maybe there is a way we can use the migration_context which we have to lookup the migration object?"
19:31:50 sean-k-mooney mriedem: hehe ya i looked at that very breifly in v1 https://review.opendev.org/#/c/667177/1/nova/virt/libvirt/driver.py@9138 and didnt foolow up as i didnt thinkwe initally needed too until artom looked at the logs.

Earlier   Later