| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-09-19 | |||
| 08:16:21 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Fix unit of hw_rng:rate_period https://review.opendev.org/683051 | |
| 08:16:56 | kashyap | mnaser: There's a lot scrollback; let's discuss today when you're around. :-) | |
| 08:22:46 | bauzas | gibi: congrats for your visa, I got mine too yesterday by postal service | |
| 08:22:58 | gibi | bauzas: congrats for you too then | |
| 08:23:00 | bauzas | (and I understand why they asked me my passport) | |
| 08:24:53 | bauzas | I have a powerful passport, it was the first time I asked for this but I was a bit afraid my visa application could be not supported because I had some coffee stains | |
| 08:25:03 | bauzas | (on my passport) | |
| 08:25:57 | kashyap | mnaser: But a couple of things to get out of the way: | |
| 08:26:15 | kashyap | mnaser: * When you use 'host-model', you won't actually see 'host-model' in the *live* guest XML; libvirt expands it (see next point) | |
| 08:26:26 | kashyap | mnaser: * To refresh your memory, 'host-model' means: (a) get maximum possible CPU features from the host; (b) make the guest live-migratable; and (c) auto-add critical CPU flags -- provided you have the right underlying bits for microcode, kernel, etc. | |
| 08:26:50 | kashyap | mnaser: * The "IvyBridge-IBRS" model does *not* contain 'vmx'; you have to explicitly add it. (You can `grep` the 'cpu_map.xml'; or if you're running new enough libvirt `grep vmx /usr/share/libvirt/cpu_map/x86_IvyBridge-IBRS.xml`) | |
| 08:30:33 | kashyap | mnaser: And lastly, since you're in the nested setup: make sure that your level-1 guest explicitly has 'vmx' flag set, even if you're using 'host-model'. | |
| 09:01:20 | bauzas | grraaah, got swamped by internal escalation | |
| 09:04:37 | openstackgerrit | Arthur Dayne proposed openstack/nova master: Fix block disk attachment failure https://review.opendev.org/682772 | |
| 09:31:35 | openstackgerrit | Boxiang Zhu proposed openstack/nova master: Preserve UEFI NVRAM variable store https://review.opendev.org/621646 | |
| 10:22:22 | openstackgerrit | Luyao Zhong proposed openstack/nova master: doc: attaching virtual persistent memory to guests https://review.opendev.org/680300 | |
| 11:07:44 | sean-k-mooney | bauzas: ya i coudl do that. i am only testing the regression cases so i would have to add more test that are not based on the conf option to make it a propper config-drive test suite | |
| 11:09:11 | sean-k-mooney | bauzas: i originaly had the tests in the test_server.py module and mriedem suggested making it a regression so i coudl avoid the old legacy code and use teh integrated helpers but since then i have added alot more tests | |
| 11:11:03 | sean-k-mooney | bauzas: so ya it definetrly worth considering factoring it into its own module and addign the remiaing tests to validate other config drive behavior e.g. passing the flage to server create and i think there is an image property you can set. so if i added 2 or 3 more tests it would fully cover the behavior | |
| 11:21:12 | mnaser | kashyap: yeah it seems we've pretty much came to the same conclusion with sean-k-mooney regarding vmx yesterday. I appreciate you actually validating our theories | |
| 11:21:25 | mnaser | Knowing libvirt expands it is good to knwo | |
| 11:21:36 | kashyap | mnaser: Yes, that's something the casual user might miss: | |
| 11:21:47 | kashyap | mnaser: "I specified 'host-model' on guest boot, why don't I see it in live?" | |
| 11:21:55 | sean-k-mooney | might miss. consider a bug... | |
| 11:21:58 | kashyap | mnaser: ... not realizing that libvirt expands it at run-time | |
| 11:22:22 | sean-k-mooney | that said i know why it does it | |
| 11:22:29 | kashyap | sean-k-mooney: Guess I broke your long sentence, sorry | |
| 11:22:48 | sean-k-mooney | not really | |
| 11:22:56 | kashyap | Ah! | |
| 11:23:09 | kashyap | You actually were commemting at my remark. (I already forgot I said "might miss") | |
| 11:23:14 | sean-k-mooney | i was just making the point that while i know why it expands it some might consier some of the expantion a bug | |
| 11:23:15 | kashyap | sean-k-mooney: Why bug? | |
| 11:23:25 | kashyap | sean-k-mooney: Ah, yeah, I see what you mean, though | |
| 11:23:31 | sean-k-mooney | because in this case its not additive | |
| 11:23:36 | kashyap | sean-k-mooney: It needs to documented clearly. | |
| 11:23:46 | kashyap | But I agree with what you imply | |
| 11:23:53 | sean-k-mooney | it replace the requested model with what was equvalent on that host | |
| 11:24:04 | kashyap | Yes | |
| 11:24:12 | sean-k-mooney | it would be require to do the current expantion if it move the vm to another host | |
| 11:24:34 | kashyap | I can't parse your second q. | |
| 11:24:47 | kashyap | s/q./comment/ | |
| 11:25:07 | sean-k-mooney | if you were to live migrate it could not rely on the models being the same | |
| 11:25:28 | sean-k-mooney | it can check but it simpler to expand the flags for a migration | |
| 11:25:46 | sean-k-mooney | so rather then specail case that they made it the default behavior | |
| 11:26:32 | sean-k-mooney | fortunetly however you added the extra_flags config | |
| 11:26:34 | kashyap | sean-k-mooney: Ah, yes; thanks for expanding your comment | |
| 11:26:54 | sean-k-mooney | so you can just add vmx or whatever you thing is missing | |
| 11:26:55 | kashyap | Yeah. sean-k-mooney -- But I haven't had the chance to thoroughly test the multiple CPU models thing we merged | |
| 11:27:03 | kashyap | Yeah | |
| 11:27:12 | sean-k-mooney | well mnaser is still on stien | |
| 11:27:22 | sean-k-mooney | at least until train is actully released | |
| 11:27:54 | sean-k-mooney | mnaser: im expecting to hear vexxhost is fully in production with train in about a week or two right :P | |
| 11:32:11 | kashyap | sean-k-mooney: If you addressed to me something after "at least until train ..." please re-post; I got dropped briefly | |
| 11:32:44 | kashyap | I just said that I realize mnaser is on Stein; but I was just talking from the POV of functional testing in Train | |
| 11:33:24 | sean-k-mooney | kashyap: no i was just admireing/makeing fun of the fact that vexxhost/osa has good ci/cd and unlike most operators actully runs a recent openstack release | |
| 11:33:46 | kashyap | Ah, nod :-) | |
| 11:33:49 | openstackgerrit | Merged openstack/nova master: libvirt: '_get_(v|p)cpu_total' to '_get_(v|p)cpu_available' https://review.opendev.org/672693 | |
| 11:34:10 | mnaser | Train on release day would be cool, we're seeing.. | |
| 11:35:40 | sean-k-mooney | osa at least in the passed did not aim to have upgrade support day 1 corret. so that will be one of the requirements pre upgrading | |
| 11:36:15 | sean-k-mooney | although you mentioned in teh project in denver that ye were tyring to get upgrade testing in the gate to run on each patch | |
| 11:36:22 | sean-k-mooney | so maybe that has changed? | |
| 11:37:24 | mnaser | We have upgrade jobs now that upgrade from Stein to master | |
| 11:37:43 | mnaser | Which afaik included placement reshuffle | |
| 11:39:02 | openstackgerrit | Merged openstack/nova master: objects: Add 'InstanceNUMATopology.cpu_pinning' property https://review.opendev.org/680106 | |
| 11:39:44 | sean-k-mooney | mnaser: cool i hope the placement work is finished in tripleo by now | |
| 12:05:15 | bauzas | mmm, super weirdo : http://paste.openstack.org/show/777697/ | |
| 12:05:29 | bauzas | but https://docs.openstack.org/api-ref/compute/?expanded=evacuate-server-evacuate-action-detail,update-forced-down-detail#update-forced-down tells me it's right | |
| 12:05:46 | bauzas | gibi: any ideas ? ^ | |
| 12:05:53 | bauzas | or sean-k-mooney: ^ | |
| 12:06:53 | bauzas | 'host' and 'binary' are correct arguments for force-down (at least for 2.77 microversion) | |
| 12:07:30 | bauzas | I don't see why I'm getting a HTTP400 | |
| 12:08:07 | gibi | looking | |
| 12:08:32 | bauzas | thanks | |
| 12:08:58 | gibi | Starting with microversion 2.53 this API is superseded by PUT /os-services/{service_id}. | |
| 12:09:39 | bauzas | ah | |
| 12:09:58 | bauzas | I remember that but the API docs don't tell it AFAICS | |
| 12:10:17 | gibi | s | |
| 12:10:17 | bauzas | holy shit, it does | |
| 12:10:17 | gibi | I copied this note from the API doc | |
| 12:10:23 | bauzas | yeah, me facepalms | |
| 12:10:49 | bauzas | FWIW, I just used some existing helper method, but I guess it sticks with an old microversion call then | |
| 12:11:02 | gibi | yeah, could be | |
| 12:11:16 | bauzas | it does indeed, 2.14 | |
| 12:11:28 | bauzas | we should deprecate this caller | |
| 12:11:59 | bauzas | and tell to only use with microversion less than 2.53 (or force an older microversion) | |
| 12:15:32 | sean-k-mooney | huh so that is what that looks like when you uses the wrong microversion | |
| 12:17:41 | bauzas | indeed, but I had to RTFM | |
| 12:17:55 | bauzas | (which I did, but my pink eyes got me wrong apparently) | |
| 12:18:10 | bauzas | so, nothing to change from an API perspective IHM | |
| 12:18:12 | bauzas | IMHO | |
| 12:18:34 | bauzas | it's just the helper method which would ideally add some docstring | |
| 13:12:43 | openstackgerrit | Merged openstack/nova master: make config drives sticky bug 1835822 https://review.opendev.org/669738 | |
| 13:12:44 | openstack | bug 1835822 in OpenStack Compute (nova) "vms loose acess to config drive with CONF.force_config_drive=True after hard reboot" [Medium,In progress] https://launchpad.net/bugs/1835822 - Assigned to sean mooney (sean-k-mooney) | |
| 13:13:34 | sean-k-mooney | TIL things can merge on there 1st time in the gate queue | |
| 13:13:41 | sean-k-mooney | is suspcious | |
| 13:22:37 | mordred | sean-k-mooney: yeah. my first thought when something is green the first time is that I did something wrong | |
| 13:23:36 | sean-k-mooney | i think we are finally getting past the gate instablity and have nailed down some of the intermitly failing tests | |
| 13:24:03 | sean-k-mooney | there are still more work to be done but its looking a lot healtier then it was on monday | |
| 13:29:18 | mordred | \o. | |
| 13:29:20 | mordred | gah | |