| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-08-29 | |||
| 14:14:11 | Roamer` | so, yeah, anybody up for reviewing a trivial "add io=native to the StorPool volume driver" change? :) https://review.opendev.org/#/c/676172/ | |
| 14:14:14 | sean-k-mooney | bauzas: what would | |
| 14:14:19 | spatel | I have tried to create fix-ip with neutron but its not letting me do that because of policy issue. | |
| 14:14:24 | bauzas | sean-k-mooney: given people were using vcpu_pin_set for pinning, we should stick with only VCPU in this case | |
| 14:14:40 | sean-k-mooney | bauzas: no that breaks upgrades | |
| 14:14:50 | bauzas | gosh | |
| 14:15:09 | sean-k-mooney | vcpu_pin_set defines the set of host cpu that can be used for pinned or floating vms | |
| 14:15:11 | mriedem | stephenfin: i never saw any operator replies to "More upgrade issues with PCPUs - input wanted" - i wonder if you should re-post that but tag with [nova] ? | |
| 14:15:17 | bauzas | sean-k-mooney: what breaks upgrades is having less VCPU between two releases with exact same config options | |
| 14:15:27 | stephenfin | sean-k-mooney: I think you're getting mixed up here | |
| 14:15:30 | mriedem | people, including myself, filter non-tagged things into a general folder and might have missed it b/c of all the noise | |
| 14:15:45 | stephenfin | mriedem: Good call. I need to reply with my resolution for that anyway | |
| 14:15:52 | artom | mriedem, aha | |
| 14:15:52 | artom | <domain type='qemu'> | |
| 14:15:56 | bauzas | sean-k-mooney: upgrading to Train with no config change should stick with exact same reporting | |
| 14:16:00 | artom | libvirtError: Requested operation is not valid: cpu affinity is not supported | |
| 14:16:07 | sean-k-mooney | bauzas: yes | |
| 14:16:10 | artom | So looks like the virt_type=kvm config didn't take | |
| 14:16:12 | sean-k-mooney | well no | |
| 14:16:12 | stephenfin | sean-k-mooney, bauzas: While 'vcpu_pin_set' is configured, we continue to *only* report VCPU | |
| 14:16:15 | sean-k-mooney | it should not | |
| 14:16:17 | sean-k-mooney | and cannot | |
| 14:16:23 | stephenfin | and the exact same amount of them as before | |
| 14:16:30 | mriedem | artom: virt_type=kvm on the controller node https://zuul.opendev.org/t/openstack/build/095bd2d0dff74761a946c29ac69b607a/log/logs/etc/nova/nova-cpu.conf.txt.gz | |
| 14:16:31 | bauzas | but we're deprecating it, so we're cool | |
| 14:16:31 | sean-k-mooney | you did not need to define vcpu_pin_set to use pinning | |
| 14:16:44 | ivve | hey, is there any way to change chipset for kvm/qemu instances from default PIIX3 into ICH9 ? would that be done in nova.conf? | |
| 14:16:51 | mriedem | artom: it's the subnode that is misconfigured https://zuul.opendev.org/t/openstack/build/095bd2d0dff74761a946c29ac69b607a/log/logs/subnode-2/etc/nova/nova-cpu.conf.txt.gz | |
| 14:16:54 | stephenfin | We will only report PCPU if you set '[compute] cpu_dedicated_set' | |
| 14:16:55 | mriedem | still has virt_type=qemu | |
| 14:17:01 | bauzas | stephenfin: excellent then | |
| 14:17:06 | bauzas | stephenfin: re-reading https://review.opendev.org/#/c/671793/16/nova/virt/libvirt/driver.py | |
| 14:17:27 | sean-k-mooney | stephenfin: and we will have the global config to contol the schdluer translation right | |
| 14:17:41 | stephenfin | sean-k-mooney: Yup, that's here https://review.opendev.org/#/c/671801/32/nova/conf/workarounds.py | |
| 14:17:58 | artom | mriedem, I guess line 1 of https://review.opendev.org/#/c/678887/13/playbooks/legacy/nova-live-migration/run.yaml is at fault | |
| 14:18:33 | mriedem | artom: why? | |
| 14:18:35 | sean-k-mooney | that needs to be set to true | |
| 14:18:49 | artom | mriedem, I'm assuming "primary" just means the controller | |
| 14:19:01 | sean-k-mooney | stephenfin: we have to have the transaltion disabeld by default | |
| 14:19:01 | stephenfin | sean-k-mooney: This is the important thing that has changed https://review.opendev.org/#/c/671800/22/nova/virt/libvirt/driver.py@6649 | |
| 14:19:36 | mriedem | artom: oh that line 1 - hosts: primary yeah | |
| 14:19:40 | mriedem | sec, i'm going to take the easy route | |
| 14:19:57 | artom | mriedem, you're handling it? OK, I'll stay off then | |
| 14:20:03 | artom | (Thanks!) | |
| 14:20:04 | stephenfin | sean-k-mooney: I'm not sure about that. We didn't do this for the consoleauth workaround option | |
| 14:20:34 | artom | mriedem, I think just adding another task with hosts: all that just sets the virt type would work? | |
| 14:20:48 | bauzas | stephenfin: sean-k-mooney: the key thing for a successful upgrade is basically : don't change anything if you upgrade without changing config defaults | |
| 14:21:26 | bauzas | stephenfin: sean-k-mooney: so the workaround necessarly has to be activated manually before Train to U | |
| 14:21:45 | bauzas | I mean, after installing Train, and some time after it | |
| 14:21:57 | bauzas | but before upgrading to U | |
| 14:22:10 | sean-k-mooney | stephenfin: on upgrade it needs to keep working as it did before with 0 config chagnes | |
| 14:22:10 | bauzas | (do we have a name btw.) ? | |
| 14:22:20 | sean-k-mooney | so it has to be disable by defualt | |
| 14:22:21 | stephenfin | bauzas: So you're saying we should disable translation by default too? | |
| 14:22:32 | bauzas | stephenfin: the default has to | |
| 14:22:35 | openstackgerrit | Matt Riedemann proposed openstack/nova master: DNM: Run LM integration tests with NUMA flavor https://review.opendev.org/678887 | |
| 14:22:42 | bauzas | stephenfin: that doesn't mean operators won't get it | |
| 14:22:46 | mriedem | artom: hopefully that does it | |
| 14:22:52 | bauzas | stephenfin: but they'll have to opt-in | |
| 14:22:54 | sean-k-mooney | in triplo we will enable the traslation by defualt for new installs and leave it deiables for upgrades | |
| 14:23:04 | stephenfin | Damn, okay, so that has to change. I was taking what '[workarounds] enable_consoleauth' did as a basis | |
| 14:23:10 | stephenfin | that was opt-in | |
| 14:23:17 | Roamer` | ivve, is hw_machine_type what you're looking for? | |
| 14:23:23 | sean-k-mooney | stephenfin: well this is why i was also saying it should not be in workaround | |
| 14:23:32 | bauzas | stephenfin: I was about to -1 :D | |
| 14:24:06 | artom | mriedem, while that runs I'll continue faffing about with the func test | |
| 14:24:10 | artom | Thanks again | |
| 14:24:22 | stephenfin | bauzas: Leave comments though | |
| 14:24:50 | stephenfin | I won't respin now. Better to wait til more of the series has been reviewed | |
| 14:24:59 | ivve | Roamer`: yes but im kinda wondering how to set the metadata | |
| 14:25:12 | ivve | or set it in nova.conf, guessing [libvirt] stanza | |
| 14:25:16 | bauzas | stephenfin: just did :p | |
| 14:25:43 | bauzas | stephenfin: okay, so back to the bottom patch, I'll need to check that everything keeps the same | |
| 14:25:52 | bauzas | unless you play with the new option | |
| 14:26:04 | stephenfin | that doesn't kick in til near the end | |
| 14:26:17 | stephenfin | it's a scheduler-only option | |
| 14:26:19 | ivve | nova.conf would be my preference so i don't need image or flavor metas | |
| 14:27:26 | Roamer` | ivve, according to https://docs.openstack.org/nova/stein/configuration/config.html it belongs in the libvirt section of nova.conf, yes | |
| 14:28:42 | openstackgerrit | Edward Hope-Morley proposed openstack/nova stable/rocky: Force refresh instance info_cache during heal https://review.opendev.org/679271 | |
| 14:30:23 | ivve | hw_machine_type=x86_64=q35 seems to be it | |
| 14:32:19 | sean-k-mooney | ivve: you can set the machine type in the nova.conf but it can also be set in the image | |
| 14:33:02 | sean-k-mooney | https://docs.openstack.org/nova/latest/configuration/config.html#libvirt.hw_machine_type | |
| 14:33:18 | openstackgerrit | Merged openstack/python-novaclient master: Microversion 2.77: Support Specifying AZ to unshelve https://review.opendev.org/665136 | |
| 14:33:46 | ivve | thanks.. seems to be quite a few to choose from :) | |
| 14:34:14 | openstackgerrit | Stephen Finucane proposed openstack/nova master: Follow-up for I491eb7abe1405b27c552cdef142c688a46e914f2 https://review.opendev.org/679273 | |
| 14:34:16 | sean-k-mooney | i normally use the unversioned ones | |
| 14:34:31 | sean-k-mooney | so just set it to "pc" or "q35" | |
| 14:35:42 | sean-k-mooney | you can set it to s specific versioned machine type also but it is normally not required on x86 | |
| 14:36:14 | kashyap | ivve: There's a good reason to set a versioned machine type: | |
| 14:36:50 | dougsz | Has anyone seen live-migrate fail on Rocky with an empty call to placement? ` Placement API returning an error response: JSON does not validate: {} does not have enough properties` | |
| 14:36:56 | kashyap | ivve: (Typically done by installer tools.) A specific versioned machine type across Compute nodes will guarantee back-n-forth live migration. | |
| 14:37:03 | openstackgerrit | Edward Hope-Morley proposed openstack/nova stable/queens: Force refresh instance info_cache during heal https://review.opendev.org/679274 | |
| 14:37:58 | ivve | kashyap: got it, so a least common denominator just as with cpu_models then? | |
| 14:38:07 | openstackgerrit | Artom Lifshitz proposed openstack/nova master: DNM: Run LM integration tests with NUMA flavor https://review.opendev.org/678887 | |
| 14:38:14 | kashyap | ivve: Yep | |
| 14:38:55 | ivve | kashyap: im guessing using virsh capabilities to find out what hosts have available to them is the way to go? | |
| 14:39:13 | kashyap | ivve: To find out machine type? | |
| 14:39:21 | ivve | ye available types | |
| 14:39:48 | ivve | i.e. highest available, lower would obviously be available i guess | |
| 14:40:12 | ivve | as with cpu models / flags | |