| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2025-03-10 | |||
| 14:16:41 | dansmith | right, it's doing both | |
| 14:17:28 | dansmith | gibi: this is the "don't do it here because..." feedback I wanted before going further, but if you need to dig in deeper, feel free to punt until you have time to do proper thinking | |
| 14:20:10 | gibi | in short, there is allocation healing in that code path which will go away. There is inventory generation there which remains there. So extending the inventory generation with this feature might be OK. | |
| 14:20:34 | gibi | so maybe it is OK | |
| 14:20:47 | dansmith | okay | |
| 14:20:49 | gibi | independenty I'm not sure how racy this is with the pci_claim | |
| 14:21:06 | dansmith | the pci claim on the compute node happens before this | |
| 14:21:07 | gibi | if this happens independently from the pci_claim in the periodic then we have a race window | |
| 14:21:38 | dansmith | the benefit here is that if you mark a device as OTU after it has already been allocated to an instance, it will be "healed" to reserved state | |
| 14:21:44 | dansmith | s/the/a benefit/ | |
| 14:21:47 | gibi | I see | |
| 14:21:52 | dansmith | I'm not sure what race you mean, | |
| 14:22:12 | dansmith | there's a small window where we could fail before we do the reserve, but if that happens, the instance will not boot | |
| 14:22:17 | dansmith | so it's safe and the window is very small | |
| 14:22:25 | dansmith | I don't think there's a race | |
| 14:22:45 | dansmith | this happens synchronously at the end of the instance_claim() method | |
| 14:22:52 | dansmith | (*and* in the periodic) | |
| 14:22:57 | gibi | I mean the pci claim claims an OTU device and returns to the compute manager spawn a VM and a periodic update_available_resouces runs after a while and sees the allocated OTU device and reserve it | |
| 14:23:10 | dansmith | no, | |
| 14:23:13 | dansmith | it happens synchronously | |
| 14:23:20 | gibi | if this happens in the same code path as the claim then cool | |
| 14:23:28 | dansmith | this doesn't depend on the periodic running before it's reserved | |
| 14:23:32 | gibi | OK | |
| 14:23:40 | gibi | that was basically my second question | |
| 14:23:44 | dansmith | runs right here in instance_claim() https://github.com/openstack/nova/blob/276685b3db6e8f2ad59c33bc254461c255700ff8/nova/compute/resource_tracker.py#L215 | |
| 14:24:23 | dansmith | that _update() all calls the _update_to_placement() as its first thing, which your first link above | |
| 14:25:47 | sean-k-mooney | dansmith from my perspective anything after self.pci_tracker.claim_instance in that function is ok, that is the earise that doing the reserved=total make sense | |
| 14:25:49 | dansmith | if that all seems reasonable now, I can document all the above details in the spec more thoroughly so that review is easier | |
| 14:26:01 | dansmith | sean-k-mooney: cool | |
| 14:26:12 | sean-k-mooney | althought there are ways we can fail later so it could be later | |
| 14:26:36 | gibi | dansmith: OK I see it now, no race | |
| 14:26:45 | dansmith | sean-k-mooney: I think there's always going to be some compromise, but it seems like when we're doing the pci accounting is the right balance to me | |
| 14:30:00 | sean-k-mooney | dansmith: so what your propsoing i think will work for spawn and unshleve and evacuate | |
| 14:30:29 | sean-k-mooney | i suspect self._update is reusted for cold migration/resize | |
| 14:30:46 | sean-k-mooney | ya it is here https://github.com/openstack/nova/blob/276685b3db6e8f2ad59c33bc254461c255700ff8/nova/compute/resource_tracker.py#L378 | |
| 14:31:08 | sean-k-mooney | so i think that is fine too | |
| 14:31:24 | sean-k-mooney | live migration is out os scope for this this cycle? | |
| 14:33:19 | sean-k-mooney | live migration does use a move claim but the claiming of the device in the nova db happens a littel earlier | |
| 14:34:06 | sean-k-mooney | i think its fine to still do the reserved=total in the same place, im just asking for the scope fo testing/support | |
| 14:34:53 | dansmith | I'll track down the live case, but yeah the cold move case is also good | |
| 14:35:32 | sean-k-mooney | so for cold migration i think we just need to document that the device is burned once we have got to verify_resize | |
| 14:35:36 | dansmith | I think I'm fine with live being out of scope here, especially since we can mark a device as live_migratable=false now, but there's no reason this shouldn't work the same I think, even if we have to do it in a slightly different place on the dest | |
| 14:35:51 | dansmith | sean-k-mooney: yeah once you've touched both devices on either end, both are burned :) | |
| 14:36:21 | sean-k-mooney | exactly, some opt might assuem that happens at confirm | |
| 14:36:25 | sean-k-mooney | so we just need to document that | |
| 14:36:38 | opendevreview | Merged openstack/nova master: Update conductor and filters allowing migration with SR-IOV devices https://review.opendev.org/c/openstack/nova/+/942146 | |
| 14:37:46 | dansmith | sure | |
| 14:40:20 | gibi | dansmith: yeah by default flavor based PCI devices are not live migratable, only exception is via the recently added live_migratable flag in the dev_spec and alias but we don't expect that will be set for OTU devices | |
| 14:40:46 | dansmith | oh, okay well, that's even easier I guess | |
| 14:41:00 | sean-k-mooney | ya the default helps | |
| 14:41:06 | gibi | only neutron based PCI devs are live migrated by default via unplug and plug | |
| 14:41:18 | sean-k-mooney | right but we dont support them in placment yet | |
| 14:41:22 | gibi | yes | |
| 14:41:25 | sean-k-mooney | so we can also ignor them | |
| 14:41:29 | sean-k-mooney | we could add a check | |
| 14:41:29 | gibi | yes | |
| 14:41:39 | sean-k-mooney | that you dont set otu and physical_network | |
| 14:41:56 | sean-k-mooney | since we know thats invalid today | |
| 14:42:09 | sean-k-mooney | but its not strictly required | |
| 14:42:11 | opendevreview | Gaudenz Steinlin proposed openstack/nova master: Make SSH compression for remote copy configurable https://review.opendev.org/c/openstack/nova/+/943377 | |
| 14:44:41 | dansmith | Okay well, I'll add some notes to the spec for this to cover the above details | |
| 15:06:03 | opendevreview | Sylvain Bauza proposed openstack/nova master: doc: mark the maximum microversion for 2025.1 Epoxy https://review.opendev.org/c/openstack/nova/+/943948 | |
| 15:10:22 | opendevreview | Dan Smith proposed openstack/nova-specs master: Add one-time-use-devices spec https://review.opendev.org/c/openstack/nova-specs/+/943486 | |
| 15:11:37 | gibi | dansmith: left couple of comments but I think the direction looks good | |
| 15:11:49 | dansmith | gibi: thanks! | |
| 15:12:34 | gibi | the only think I'm a bit hesitant about it is the long term effect of the decision that we go with placement to solve this while and therefore we are forced not to support OTU VFs. | |
| 15:13:28 | dansmith | gibi: tbh, I'm not really sure when VFs would need to use this, as I suspect any virt-knowing device would be sensitive to exposure between uses (perhaps not) | |
| 15:14:07 | dansmith | that said, I think that from the operator point of view, it doesn't necessarily matter that it's happening in placement (other than the steps to unreserve after cleaning I guess).. but the devspec wouldn't need to change if we started accounting for the one-time-use in, say, PCI tracker | |
| 15:14:20 | gibi | yeah that is a valid point. This is why I'm not against going with the placement route, just hesitant that our udnerstanding of SRIOV VFs are correct | |
| 15:14:53 | gibi | dansmith: yeah the unreserve is the problematic step if we go with the PCI tracker route | |
| 15:14:57 | dansmith | right | |
| 15:15:01 | gibi | nova has no API for that | |
| 15:15:16 | dansmith | an SRIOV NIC is safe for use without cleaning, because you can't update firmware I think | |
| 15:15:46 | dansmith | not sure about a VGPU and memory exposure I guess, but .. they have to know that the point of that is sharing, so I would expect it's okay | |
| 15:15:52 | gibi | I also think that, but I never checked what happens with GPU VFs or NVMe VFs | |
| 15:16:27 | dansmith | an SRIOV NVMe would be a problem I guess, but if you're de- and re-allocating them they should be safe (like new namespaces on a regular one) | |
| 15:17:03 | gibi | true, if the VF is mapped to a namespace then it is OK | |
| 15:17:06 | dansmith | yeah | |
| 15:17:26 | gibi | OK. I'm less hesitant now. We have good reasons to believe that this feature is not needed for VFs | |
| 15:17:46 | gibi | a sane VF implementor needs to handle the sharing issue | |
| 15:17:51 | dansmith | I think we just make sure it's clear in the doc that this is only for whole devices because they're not virt-aware, and we can also say "for more comprehensive situations you need cyborg anyway" :) | |
| 15:17:55 | dansmith | yeah | |
| 15:18:21 | gibi | OK | |
| 15:18:36 | gibi | and :) for cyborg as a fallback | |
| 15:19:42 | dansmith | I really don't want to blur the line of nova-vs-cyborg much, so I think keeping this limited to specific uses is helpful to avoid people trying to push nova to do more | |
| 15:19:49 | opendevreview | Sylvain Bauza proposed openstack/nova master: Update compute rpc alias for epoxy https://review.opendev.org/c/openstack/nova/+/943952 | |
| 15:19:57 | sean-k-mooney | dansmith: gibi you cant update teh frimware vai the vf from an nic i have worked with | |
| 15:20:23 | dansmith | right | |
| 15:20:24 | sean-k-mooney | dansmith: you can flash ab bitrstream via a specific VF on an fpga that change the behvior of other VFs | |
| 15:20:35 | sean-k-mooney | but that a very different usecasue | |
| 15:20:46 | sean-k-mooney | to me that is not what you are creating this for | |
| 15:20:52 | dansmith | definitely not | |
| 15:20:59 | sean-k-mooney | if we add pci groups in the future | |
| 15:21:07 | sean-k-mooney | we could extend the concpet to the group | |
| 15:21:17 | sean-k-mooney | i.e. a one time group | |
| 15:21:52 | dansmith | tbh, the pci-in-placement approach seems to me to be a bit of a stretch of the model, since placement really assumes resources of a given class are fungible and does not track actual assignment | |
| 15:22:23 | sean-k-mooney | dansmith: well in general tehy are fungible | |
| 15:22:24 | dansmith | so anything specific to a given VF probably needs a new model and new API (which cyborg is well-positioned to handle, IMHO) | |
| 15:22:46 | dansmith | sean-k-mooney: VFs are, but not whole PCI devices | |
| 15:23:17 | sean-k-mooney | maybe you could argue it either way | |