| opendevreview | Takashi Kajinami proposed openstack/diskimage-builder master: Remove workaround for missing ndisc6 in EPEL8 https://review.opendev.org/c/openstack/diskimage-builder/+/926458 | 03:56 |
|---|---|---|
| opendevreview | Benjamin Schanzel proposed zuul/zuul-jobs master: Add ensure-dib role https://review.opendev.org/c/zuul/zuul-jobs/+/922910 | 13:26 |
| opendevreview | Benjamin Schanzel proposed zuul/zuul-jobs master: Add build-diskimage role https://review.opendev.org/c/zuul/zuul-jobs/+/922911 | 13:26 |
| opendevreview | Benjamin Schanzel proposed zuul/zuul-jobs master: Add build_diskimage_environment role variable https://review.opendev.org/c/zuul/zuul-jobs/+/926224 | 13:26 |
| opendevreview | Benjamin Schanzel proposed zuul/zuul-jobs master: Add a diskimage-builder job https://review.opendev.org/c/zuul/zuul-jobs/+/926225 | 13:26 |
| opendevreview | Takashi Kajinami proposed openstack/diskimage-builder master: Install ndisc6 using pkg-map https://review.opendev.org/c/openstack/diskimage-builder/+/926458 | 14:39 |
| fungi | huh... let's encrypt e-mailed us to let us know that we have a cert for openmetal.us-east.opendev.org expiring in 19 days | 18:03 |
| fungi | and i have just confirmed that's the current expiration on the cert https://openmetal.us-east.opendev.org/ is using | 18:04 |
| fungi | i guess we're managing that by hand for now, not via our usual ansible? | 18:04 |
| clarkb | fungi: kolla set it up. I wonder if we need to do something similar there to what we did with the old linaro cloud | 18:08 |
| clarkb | basically trigger kolla to do things to refresh it | 18:08 |
| clarkb | frickler may know how that is expected to work in kolla | 18:08 |
| frickler | I was hoping kolla would be that on its own, but maybe it needs a trigger | 18:08 |
| fungi | we have a couple of weeks to work it out | 18:08 |
| frickler | I can check next week | 18:08 |
| fungi | yeah, there's no rush | 18:08 |
| frickler | also it seems we should add this to our certcheck list? | 18:09 |
| fungi | i can do that now, yep | 18:09 |
| clarkb | ++ | 18:09 |
| opendevreview | Jeremy Stanley proposed opendev/system-config master: Track OnMetal environment's HTTPS cert expiration https://review.opendev.org/c/opendev/system-config/+/926488 | 18:13 |
| clarkb | fungi: two small things on ^ one of which is important for functionality | 18:16 |
| fungi | gah, i keep making that same typo | 18:29 |
| fungi | as for the port, yes it looks like the same cert is being used on both listening ports, i agree it would be safer to check the one we're relying on | 18:31 |
| opendevreview | Jeremy Stanley proposed opendev/system-config master: Track our OpenMetal environment HTTPS cert expiry https://review.opendev.org/c/opendev/system-config/+/926488 | 18:33 |
| johnsom | When it's a good time, I would like to talk about getting an image setup for nodepool that enables the igb driver (again). I have poked around at this and nova won't pick this setting up from the flavor, it appears to only be implemented in the image properties. | 19:21 |
| clarkb | I'm around. Just finishing up lunch. Is that a bug? The hardware driver presented to a VM seems liek a flavor attribute more than an image setting from a logical standpoint | 19:26 |
| clarkb | images ship with many many device drivers. The flavor should determine what is presented to the kernel and then udev or whatever will plug things in | 19:26 |
| johnsom | I don't disagree, but it seems that almost all of the qemu configuration settings are image properties and not "extra_specs" in the flavor | 19:26 |
| clarkb | the first thing we need to determine is what if any clouds that we talk to support this functionality. From memory the last time this came up I suspected that vexxhost and openmetal may | 19:28 |
| clarkb | then based on that we can figure out if we just update the images uploaded to those clouds (the different device type should work normally for regular jobs I think so we should be able to do that transparently) | 19:28 |
| clarkb | or if we need to do a whole separate image upload | 19:28 |
| johnsom | The list of settings available via image properties: https://docs.openstack.org/glance/latest/admin/useful-image-properties.html | 19:29 |
| clarkb | if the driver is expected to just work transparentyl for normal job s then ya Ithink we can update https://opendev.org/openstack/project-config/src/branch/master/nodepool/nodepool.yaml#L79-L101 maybe for one image to start and see what happens | 19:29 |
| clarkb | and if that works expand to all the images in that cloud and then rinse and repeat for othe rclouds we expect this to work in | 19:30 |
| johnsom | Yeah, it should work fine for normal jobs as well. My only "caution" is this driver is fairly new. | 19:30 |
| clarkb | right the main concern is if we se tthat flag on the wrong cloud the image will stop working in that cloud | 19:31 |
| johnsom | The trick is the hypervisor must be running Noble to have a version of qemu that supports it | 19:31 |
| clarkb | johnsom: https://zuul-ci.org/docs/nodepool/latest/openstack.html#attr-providers.[openstack].diskimages.meta that is the setting to update in the block I linked previously | 19:32 |
| clarkb | johnsom: in the caes of openmetal it is centos 9 stream I think and when I checked centos 9 stream package versions they seemed new enough to support it | 19:32 |
| clarkb | I don't know what vexxhost is running these days but they have historically kept things relatively up to date | 19:32 |
| johnsom | Yeah, that looks right | 19:32 |
| clarkb | rackspcae won't support it (no qemu) and I suspect ovh won't either | 19:33 |
| johnsom | Ah, cool. I haven't checked c9s recently to see if it was updated. It's in Fedora as well, so makes sense to be in c9s now | 19:33 |
| clarkb | so ya I would start with openmetal on a single image as a test. If networking breaks nodepool won't use the node so its relatively safe as long as we do it in a small subset of nodes | 19:33 |
| johnsom | +1, I am happy to do testing legwork on this | 19:34 |
| clarkb | then if we get it working we can figur eout labels or whatever so yo ucan select jobs run on those nodes. But we can start small and just see that it works at all | 19:34 |
| clarkb | johnsom: can you push up the change to modify that nodepool.yaml file? I suspect you've got a better grasp of what all the value sshould be in the image metadat | 19:37 |
| johnsom | Yeah, I can work on that. | 19:37 |
| johnsom | Ah... I hit a roadblock. nova hasn't added that driver to the valid vif model list yet. | 19:40 |
| johnsom | Ok, let's pause to get that fixed which means this is now pushed out a long time. | 19:41 |
| clarkb | oof | 19:45 |
| fungi | it should only take 5-6 years for most cloud providers to upgrade nova ;) | 20:21 |
| fungi | (hopefully the modern upgradeability of openstack is shortening that window more recently) | 20:21 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!