| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-04-23 | |||
| 13:17:13 | sean-k-mooney | sahid: im nologer the product owner for that ci but i can reach out internally and ask if they can enable it | |
| 13:17:52 | sean-k-mooney | infact we should proably also enable multiqueue also | |
| 13:17:57 | sahid | sean-k-mooney: that would be great, i did some internal tests, but dansmith really would like to see some ci running | |
| 13:18:00 | sahid | https://review.openstack.org/#/c/484997/ | |
| 13:18:06 | sahid | sean-k-mooney: that would be great yes | |
| 13:19:28 | hkominos | hi: in order to spawn an openstackVM nova will define an xml file describing the VM.In the XML I think nova also defines a number of pci slot for taht VM. Is that default number of PciE slots taken from some config file? | |
| 13:20:27 | hkominos | Just to clarify, I want nova to add another pci-slot to the Vms that it creates | |
| 13:21:24 | sean-k-mooney | sahid: so out of interest what was the issue that was found just before merging? | |
| 13:24:29 | sahid | libvirt rejects driver name=vhost-user | |
| 13:24:40 | sahid | only acceptable values are qemu and vhost | |
| 13:29:42 | kashyap | Yeah, it is also documented so in the formatdomain.html | |
| 13:30:05 | kashyap | (Probably the upstream doc of that sub section needs a bit more tweaking.) | |
| 13:30:58 | openstackgerrit | Margarita Shakhova proposed openstack/nova master: Do not skip migrations in _destroy_evacuated_instances() https://review.openstack.org/563623 | |
| 13:32:46 | dansmith | sean-k-mooney: does that ci not support depends-on? | |
| 13:34:06 | hkominos | bauzas: in order to spawn an openstackVM nova will define an xml file describing the VM.In the XML I think nova also defines a number of pci slot for taht VM. Is that default number of PciE slots taken from some config file? | |
| 13:35:20 | bauzas | hkominos: the libvirt driver gets the number of PCI devices by calling libvirtd and then creating a config | |
| 13:35:40 | sean-k-mooney | dansmith: that ci does not it is still using zuul 2.5 with jenkins | |
| 13:35:43 | bauzas | hkominos: then the config is used by the driver.py module to create a XML | |
| 13:36:00 | bauzas | when I say a config, I mean a config object | |
| 13:36:07 | sean-k-mooney | dansmith: well the old depends on might work | |
| 13:36:18 | bauzas | hkominos: see virt.libvirt.config | |
| 13:37:12 | sean-k-mooney | dansmith: we had planned to move it to zuul v3 at some point but not sure if the new team that is managing that intend to do that | |
| 13:37:59 | dansmith | sean-k-mooney: and that only runs on neutron, so no way to propose some sort of change that would end up with some nova config change? | |
| 13:38:30 | sean-k-mooney | dansmith: the intel nfv ci runs on changes to nova also | |
| 13:38:38 | hkominos | bauzas: thx I will take a look.Should I by intuitio,n assume that it is possible to just ask libvirt to create another empty pci slot? | |
| 13:39:25 | dansmith | sean-k-mooney: oh does it? then we could just use a DNM change to default those queues to something right? | |
| 13:39:36 | sean-k-mooney | dansmith: currently it runs 3 jobs on nova tempest-dsvm-ovsdpdk-nfv-networking-xenial tempest-dsvm-multinode-ovsdpdk-nfv-networking-xenial and tempest-dsvm-intel-nfv-xenial | |
| 13:39:54 | sean-k-mooney | dansmith: yep you could | |
| 13:39:55 | dansmith | ah yeah I see | |
| 13:40:12 | dansmith | sahid: easy ^ | |
| 13:43:01 | kashyap | What is Eric Fried's Friday nick? | |
| 13:43:18 | cdent | kashyap: usually fried_rice | |
| 13:43:38 | kashyap | cdent: Ah, thanks. I think he's AFK now. | |
| 13:44:35 | sahid | dansmith, sean-k-mooney are the repo of the template to i can provide a patch? | |
| 13:44:39 | sahid | where* | |
| 13:46:13 | cdent | kashyap: he's at a conference this week | |
| 13:46:38 | dansmith | sahid: just nova.. I'll do it, just a sec | |
| 13:47:54 | kashyap | cdent: Noted; thanks. | |
| 13:48:17 | openstackgerrit | Dan Smith proposed openstack/nova master: DNM: Test virtio queue sizes https://review.openstack.org/563626 | |
| 13:48:23 | dansmith | sahid: ^ | |
| 13:49:06 | sahid | dansmith: oh... smart :) | |
| 13:49:47 | cdent | scheduler meeting #openstack-meeting-alt in ten minutes | |
| 13:49:53 | bauzas | hkominos: well, not sure I understant your question, you're asking if Nova should provide an empty pci slot anyway? | |
| 13:50:35 | kashyap | As a second pair of eyes to MattR, can anyone else be able to take a look at this: https://review.openstack.org/#/c/558783/ | |
| 13:53:35 | hkominos | bauzas:yes | |
| 13:54:13 | hkominos | bauzas:I think it does provide one anyway but I need one more. | |
| 13:54:23 | bauzas | hkominos: well, I don't know why, so you should ask for that by a blueprint | |
| 13:54:28 | tssurya | dansmith: if I want to populate a newly added column (say queued_for_delete), based on existing data from another column (say instances.deleted), what is the best way of doing this ? -> migration tool (seems to be the usual way in the past) ? or is there a way to do this while adding the new column ? | |
| 13:55:01 | dansmith | tssurya: doing it while adding the column should not be done because it can lock the table and be a problem for pre-applying schema | |
| 13:55:17 | dansmith | tssurya: if you can, the best way is on loading/saving the object out of the db | |
| 13:55:48 | dansmith | tssurya: so in _from_db_obj or something like that, set the field on the loaded object so it will be saved later, optionally saving it right then if you need it to be saved immediately | |
| 13:55:59 | tssurya | dansmith: oh okay, yep will go that way then | |
| 13:56:15 | dansmith | tssurya: example: https://github.com/openstack/nova/blob/master/nova/objects/service.py#L259-L264 | |
| 13:56:22 | dansmith | that was how we generated uuids on the service objects | |
| 13:56:57 | tssurya | dansmith: thanks | |
| 13:57:19 | tssurya | I will use this approach | |
| 14:04:19 | sean-k-mooney | sahid: the nfv cis are corntolled by internal repos in intel. you wont be able to change the test setup. you will however be able to hardcode the use of a queue lenght in nova in a DNM patch and get the same effect | |
| 14:05:59 | stephenfin | bauzas: All comments addressed here. Fancy hitting it again? https://review.openstack.org/#/c/379525/ | |
| 14:06:20 | bauzas | stephenfin: yup, I can sneak peek on it | |
| 14:06:42 | sean-k-mooney | sahid: we might be able to change the setup via a patch to networking-ovs-dpdk to add a local.sh or late running section to the devstack pluging to alter the flavors used by the ci but i would need to look into that. | |
| 14:08:45 | sahid | sean-k-mooney: ok thanks | |
| 14:09:33 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add policy rule to block image-backed servers with 0 root disk flavor https://review.openstack.org/561284 | |
| 14:14:04 | kashyap | Can anyone remind me what is the way to recheck this job: 'legacy-tempest-dsvm-multinode-live-migration' | |
| 14:17:11 | kashyap | Seems like I'm hitting a problem that is related to the CI env: http://logs.openstack.org/12/562312/4/check/legacy-tempest-dsvm-multinode-live-migration/88f1334/job-output.txt.gz | |
| 14:26:14 | mriedem | kashyap: you can't recheck specific jobs | |
| 14:26:22 | mriedem | you just leave a 'recheck' comment | |
| 14:26:58 | kashyap | mriedem: Yeah, that I'm aware. Just thought I'd check if I can avoid a "full blown recheck" | |
| 14:27:16 | kashyap | (To be mindful to not cause needless CI churn.) | |
| 14:27:58 | mriedem | mnaser: i got https://review.openstack.org/#/c/561284/ updated to use a policy rule instead of a config option, this is the thing to disable booting from a 0 root disk flavor if you're not doing bfv | |
| 14:29:31 | openstackgerrit | Merged openstack/nova master: libvirt: Add a debug log entry before / after invoking migrate() https://review.openstack.org/563550 | |
| 14:29:38 | openstackgerrit | Merged openstack/nova master: Mention that users need noVNC >= 0.6 https://review.openstack.org/562847 | |
| 14:30:30 | kashyap | mriedem: When you have time this week: Zuul blessed this, and I've addressed your remarks: https://review.openstack.org/#/c/558783/ (libvirt: Bump MIN_{LIBVIRT,QEMU}_VERSION for "Rocky") | |
| 14:33:37 | mriedem | was just looking at it | |
| 14:39:34 | kashyap | Thanks. | |
| 14:40:02 | kashyap | stephenfin: efried: Thanks for the quick review (commit: 880f628)! | |
| 14:40:20 | efried | kashyap: yahyoubetcha. Easy one. | |
| 14:41:19 | kashyap | efried: More context on that is - before that commit, it was incredibly difficult to tell if that migrate() method has run or not | |
| 14:41:31 | efried | I get it. | |
| 14:42:23 | efried | kashyap: Mainly, I get credit for a libvirt review I can point to so people don't accuse me of being PowerVM/IBM-centric. | |
| 14:42:25 | kashyap | Although there's a comment snuck in to Eventlet schedule a new thread, but that's too subtle | |
| 14:42:33 | kashyap | efried: Hehe | |
| 14:43:01 | efried | kashyap: I'm counting on major brownie points from this series: https://review.openstack.org/560444 | |
| 14:43:02 | kashyap | efried: Don't worry, I'll send more your way; mechanical clean up of constants :P | |
| 14:43:20 | mnaser | mriedem: nice patch, is there a way we can track things like "things to do for next release" ? | |
| 14:43:24 | kashyap | Until you might get sick of it :P (Once Matt merges the earlier MIN_ bump) | |
| 14:43:51 | mriedem | mnaser: we don't have anything formal for that besides the TODOs in the code | |
| 14:45:14 | mnaser | i see | |
| 14:45:44 | kashyap | efried: Interesting. Have to still educate myself on that (ProviderTree stuff) | |
| 14:46:36 | efried | kashyap: Let me know if you want help there. We wrote a nice contributor doc - maybe since you claim to be fresh, you could scan it and let me know if it gives you what you need... | |
| 14:46:46 | kashyap | efried: I don't claim to be fresh :P | |
| 14:47:14 | kashyap | efried: But yeah, got a URL to that doc? | |
| 14:47:29 | efried | kashyap: working on that (I pretty much suck at searching the docs) | |
| 14:47:45 | kashyap | Aah, I see. The past tense word "wrote" made me think it's already in tree | |
| 14:48:31 | efried | kashyap: It is. | |
| 14:48:37 | efried | kashyap: Mahbad, it's a reference doc: https://docs.openstack.org/nova/latest/reference/update-provider-tree.html | |
| 14:49:17 | efried | kashyap: So yeah, in recompense for that two-line review, you get to read my doc and give me feedback. | |
| 14:49:25 | efried | kashyap: No mistake, you got the raw end of that deal. | |
| 14:50:28 | kashyap | efried: Hehe, oh on the first look, I appreciate the structure & detail. This needs a bit of concentration to read cover-to-cover; queued it. | |
| 14:50:57 | efried | kashyap: thanks in advance | |
| 14:51:59 | kashyap | efried: A quick question -- do you also do "hypervisor stuff" with PowerVM / IBM (z)? | |