Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-05-15
12:17:48 lyarwood jsuchome: https://review.opendev.org/727347 should fix the pep8 errors
12:18:20 jsuchome I see, thanks
12:39:16 openstackgerrit Andreas Jaeger proposed openstack/python-novaclient master: Switch to newer openstackdocstheme and reno versions https://review.opendev.org/728270
12:39:19 openstackgerrit Andreas Jaeger proposed openstack/nova master: Switch to newer openstackdocstheme and reno versions https://review.opendev.org/728263
12:58:57 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Update and correct typing information https://review.opendev.org/714694
12:59:33 stephenfin lyarwood: Addressed your comments on ^
13:07:45 lyarwood stephenfin: ack thanks
13:08:54 openstackgerrit Balazs Gibizer proposed openstack/nova master: Reserve DISK_GB resource for the image cache https://review.opendev.org/728439
13:09:39 gibi dansmith: hi! fyi there is the first version of disk reservation patch for the image cache ^^
13:10:45 gmann gibi: thanks
13:11:51 francoisp lyarwood, bauzas hello, when you get a chance: https://review.opendev.org/#/c/726508/ -thank you!
13:13:00 lyarwood francoisp: ack sorry done
13:15:50 francoisp thanks lyarwood
13:25:20 openstackgerrit Merged openstack/nova master: Bump hacking min version to 3.0.1 https://review.opendev.org/727347
13:33:52 openstackgerrit Jiri Suchomel proposed openstack/nova master: Bring back allowed_direct_url_schemes in support of RBD image download https://review.opendev.org/728095
13:35:48 gmann finally merged. gate should be green now
13:41:35 openstackgerrit Jiri Suchomel proposed openstack/nova master: Adapt the admin guide to describe the direct RBD export https://review.opendev.org/728411
13:44:17 dansmith gibi: you decided on reserved instead of a new allocation?
13:44:43 gibi dansmith: yes, reservation is the current working assumption
13:44:44 dansmith gibi: I really wasn't arguing that reserved was better (and I think I said that) I just wanted us to consider it
13:45:51 dansmith gibi: it certainly looks simpler than a new allocation though. I was going to look for code that deleted the allocation if they moved the cache from shared to separate filesystem, but this would auto correct itself
13:46:11 gibi dansmith: I realized that we recalculate reservation at every update_provider_tree() call anyhow
13:46:20 dansmith yeah
13:46:39 gibi so I plugged my calculation into that loop
13:47:23 dansmith yeah, even though this is less granular, it feels like a much smaller and less scary change, even something we could backport
13:47:52 gibi there is one downside, placement allow bumping the reservation over the total - usage
13:48:44 dansmith the other downside is this lags the actual usage
13:48:50 dansmith but, very simple
13:49:59 gibi when I pushed the patch I added a TODO to bump the reservation before we download an image but since then I lost in the imagebackend code again to find the place where we download image to the cache and it is faaar from easy
13:50:21 dansmith yeah, no
13:52:00 gibi I start thinking that from the reservation handling perspective this is what we can do. This is far from perfect but it has a good roi
13:53:19 gibi dansmith: would it be OK with you if we stop in the virt driver at this point? I will still look into adding the pre-filter part separately
13:53:43 gibi but I'm scared from the imagebackend
13:54:37 dansmith what do you mean "stop in the virt driver"?
13:55:49 gibi I mean, not trying to bump the reservation before we download a new image to the cache
13:57:09 dansmith oh yeah, I don't think that's the way to go.. we'd have to update inventory before each boot
13:57:50 dansmith I think if we go the reservation approach, we accept the imperfectness of it, in exchange for its simplicity
13:58:10 dansmith if it turns out to not help often enough, then we can move to the allocation method or something else
13:58:17 dansmith a pre-filter will help avoid that
13:58:41 dansmith I dumped some comments about the approaches on the patch just now
13:58:46 gibi looking
13:59:43 openstackgerrit Lee Yarwood proposed openstack/nova master: compute: Allow snapshots to be created from PAUSED volume backed instances https://review.opendev.org/728011
13:59:50 openstackgerrit Lee Yarwood proposed openstack/nova master: compute: Remove snapshot quiesce tests for STOPPED and SUSPENDED instances https://review.opendev.org/728012
14:02:18 gibi OK so we agree that this is a good compromise as is. I will move the config to [workaround] for now as that is safer place.
14:02:53 gibi an sure I will add test and documentation to the patch
14:03:01 dansmith cool
14:03:27 gibi dansmith: thank you
14:03:38 dansmith gibi: you're doing the work, but... sure :)
14:04:26 gibi dansmith: talking about it helps :)
14:04:56 dansmith gibi: my couch is always open
14:05:00 gibi :D
14:09:00 aarents gibi: dansmith thanks for handling this reservation stuff, I will backport this once finished for sure :)
14:09:38 gibi aarents: I have a mandate to backport it to Train at least due to the downstream project I'm supporintg
14:11:20 dansmith especially if it's in [workarounds] I have no problem supporting it going back to train upstream, given it's something we broke and there are reports in the wild
14:12:34 aarents gibi Ok I will need it for stein, I will backport it to stein if possible
14:12:43 gibi aarents: cool
14:14:09 dansmith I think at some point in the lineage the provider tree stuff will differ enough that it will stop being trivial
14:17:07 dansmith aarents: any chance you'd take an unmerged version of this, apply to one host and let us know how it works for you?
14:17:40 dansmith it would suck to do all this, backport it way back and then find out the small race windows still let too much stuff through
14:20:15 aarents dansmith: yes will try to do it next week
14:20:34 dansmith cool
14:23:58 gibi aarents: I try to add some test to the patch still today
14:26:21 aarents ok
14:51:38 jsuchome dansmith: once you have time ... I addressed your comments in https://review.opendev.org/#/c/574301/
14:52:05 dansmith yep, in the queue
14:52:36 jsuchome great
14:57:30 openstackgerrit Andreas Jaeger proposed openstack/python-novaclient master: Switch to newer openstackdocstheme and reno versions https://review.opendev.org/728270
15:00:55 dansmith jsuchome: btw, the spec really needs to be merged before we move too far on the code
15:02:48 openstackgerrit Andreas Jaeger proposed openstack/nova master: Switch to newer openstackdocstheme and reno versions https://review.opendev.org/728263
15:38:21 jsuchome well, I just thought I start with something... nothing I can do with the spec right now
15:39:51 openstackgerrit Stephen Finucane proposed openstack/nova master: tox: Integrate mypy https://review.opendev.org/676208
15:39:52 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Update and correct typing information https://review.opendev.org/714694
15:39:52 openstackgerrit Stephen Finucane proposed openstack/nova master: libvirt: Add typing information https://review.opendev.org/714695
15:39:53 openstackgerrit Stephen Finucane proposed openstack/nova master: tests: Split instance NUMA object tests https://review.opendev.org/714696
15:39:53 openstackgerrit Stephen Finucane proposed openstack/nova master: objects: Replace 'cpu_pinning_requested' helper https://review.opendev.org/714697
15:39:54 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Don't consider overhead CPUs for unpinned instances https://review.opendev.org/714698
15:39:54 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Remove handling of pre-Train compute nodes https://review.opendev.org/714699
15:39:55 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Add validation for 'cpu_realtime_mask' https://review.opendev.org/468203
15:39:55 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Tweak the 'cpu_realtime_mask' handling slightly https://review.opendev.org/461456
15:39:56 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Rework 'get_realtime_constraint' https://review.opendev.org/714700
15:39:56 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Invert order of NUMA topology generation https://review.opendev.org/714701
15:39:57 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Remove '_numa_fit_instance_cell_with_pinning' https://review.opendev.org/714703
15:57:04 openstackgerrit Huaqiang Wang proposed openstack/nova master: Introduce 'pcpuset' field for InstanceNUMACell https://review.opendev.org/714658
15:57:05 openstackgerrit Huaqiang Wang proposed openstack/nova master: Introduce the 'CPUAllocationPolicy.MIXED' enum https://review.opendev.org/716267
15:57:05 openstackgerrit Huaqiang Wang proposed openstack/nova master: Calculate the CPU usage for mixed instance https://review.opendev.org/716269
15:57:06 openstackgerrit Huaqiang Wang proposed openstack/nova master: online data migration for cpuset of dedicated instance https://review.opendev.org/718493
15:57:06 openstackgerrit Huaqiang Wang proposed openstack/nova master: libvirt/hardware: refactor numa_get_constraints https://review.opendev.org/718294
15:57:07 openstackgerrit Huaqiang Wang proposed openstack/nova master: Setup 'mixed' instance through 'PCPU' and 'VCPU' resource https://review.opendev.org/713355
15:57:07 openstackgerrit Huaqiang Wang proposed openstack/nova master: metadata: export the vCPU IDs that are pinning on the host CPUs https://review.opendev.org/688936
15:57:08 openstackgerrit Huaqiang Wang proposed openstack/nova master: notification: add the pcpuset info to instance numa cell payload https://review.opendev.org/728480
15:57:08 openstackgerrit Huaqiang Wang proposed openstack/nova master: libvirt: apply mixed instance CPU policy https://review.opendev.org/728481
15:57:09 openstackgerrit Huaqiang Wang proposed openstack/nova master: scheduler: mixed instance CPU resource translating https://review.opendev.org/728482
15:57:09 openstackgerrit Huaqiang Wang proposed openstack/nova master: flavor: Enable 'hw:cpu_dedicated_mask' for creating a mixed instance https://review.opendev.org/728483
15:57:10 openstackgerrit Huaqiang Wang proposed openstack/nova master: compute: bump nova-compute service version for mixed instance https://review.opendev.org/728484
15:57:10 openstackgerrit Huaqiang Wang proposed openstack/nova master: hardware: create 'mixed' instance for realtime and shared CPUs https://review.opendev.org/728485
16:05:20 openstackgerrit Stephen Finucane proposed openstack/nova master: Add '__version__' attribute https://review.opendev.org/728490
16:14:33 openstackgerrit Andreas Jaeger proposed openstack/nova master: Switch to newer openstackdocstheme and reno versions https://review.opendev.org/728263
16:17:49 openstackgerrit Balazs Gibizer proposed openstack/nova master: Reserve DISK_GB resource for the image cache https://review.opendev.org/728439
16:18:25 gibi aarents, dansmith: this time with tests ^^
16:19:25 gibi and that marks the end of my work week. Have a nice weekend o/

Earlier   Later