| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2023-06-01 | |||
| 12:56:42 | sean-k-mooney | so we could start makeing that our default memory backend in nova/libvirt | |
| 12:57:17 | sean-k-mooney | specificaly our default coudl evolve to | |
| 12:57:20 | sean-k-mooney | <memoryBacking> | |
| 12:57:22 | sean-k-mooney | <source type="memfd"/> | |
| 12:57:24 | sean-k-mooney | <access mode="shared"/> | |
| 12:57:26 | sean-k-mooney | </memoryBacking> | |
| 12:57:52 | sean-k-mooney | that will enable virtio fs and vhost-user to start workign automaticlaly without any special addtional configuration | |
| 12:58:38 | sean-k-mooney | Uggla_: however our current min qemu is 4.2.0 or declared next min is 5.2.0 | |
| 12:58:52 | sean-k-mooney | so we need to actuly raise the min version befoer we do that | |
| 12:59:21 | sean-k-mooney | kashyap was ment to be workign on that but i have not seen any patches yet | |
| 13:02:08 | Uggla_ | sean-k-mooney, as a reminder for virtiofs I require a qemu > 5.0.0 and libvirt > 6.2.0. | |
| 13:03:22 | Uggla_ | sean-k-mooney, anyway assuming I find the time to do it, any objection to write a spec for memfd ? (so we could review it at PTG ?) | |
| 13:04:06 | sean-k-mooney | go for it | |
| 13:04:27 | Uggla_ | sean-k-mooney, thx | |
| 13:04:48 | sean-k-mooney | what i have proposed in the past is adding hw:memory_backend=file|anonumous|memfd|hugepage | |
| 13:04:59 | sean-k-mooney | and if we raise the min then memfd can be the default | |
| 13:05:04 | sean-k-mooney | when its not set | |
| 13:05:43 | Uggla_ | what is anonymous ? | |
| 13:05:56 | sean-k-mooney | what we use today just speeled incorrectly | |
| 13:06:01 | sean-k-mooney | non shared 4k pages | |
| 13:07:46 | Uggla_ | ok you think there will be no "side effect" to switch to memfs as default ? | |
| 13:08:13 | sean-k-mooney | there are sideeffect but i think its a better default. provided we have a new enoguh qemu | |
| 13:08:24 | sean-k-mooney | specificly im suggeting changin two defaults | |
| 13:08:42 | Uggla_ | ok | |
| 13:09:03 | sean-k-mooney | hw:memory_backing=memfd which im suggeting should be the new default woudl set <source type="memfd"/> and <access mode="shared"/> | |
| 13:09:12 | sean-k-mooney | the <access mode="shared"/> is the main delta | |
| 13:09:34 | sean-k-mooney | memfd ensure there is a file decriptor allcoated to the memory segment | |
| 13:10:33 | sean-k-mooney | combinded with <access mode="shared"/> that means dpdk or or other tools that do security scaning tools can map the guest memory an acess it form the host | |
| 13:11:06 | sean-k-mooney | so that will enabel ovs-dpdk to work out of the box without requiring the vm to use hugepages or file backed memory | |
| 13:11:21 | Uggla_ | cool ! | |
| 13:11:41 | sean-k-mooney | and it will make virtio-fs work which is your main usecase enable but the maniall share work you are doing and the scaphanda work | |
| 13:12:19 | sean-k-mooney | the reason for keeping anonymous is for those that dotn what <access mode="shared"/> | |
| 13:12:29 | Uggla_ | yep, it might simplify the host config too. | |
| 13:13:13 | Uggla_ | sure, it good to keep anonymous for people who does not want to use shared mem, for any reason. | |
| 13:13:18 | sean-k-mooney | yes so put to gether a spec and we can debate the default in it but i would be infavor of using memfd as it has may advantages | |
| 13:14:26 | Uggla_ | I agree, will do that. | |
| 13:15:09 | Uggla_ | Another thing, I'm not sure you have seen my note yesterday about virtofs not supporting readonly. Do you thing proposing virtio9p as an alternative to virtiofs could be useful ? | |
| 13:15:41 | Uggla_ | or we should just wait virtiofs to support it ? | |
| 13:17:29 | sean-k-mooney | virtio9p does not apprer to be supported by libvirt | |
| 13:18:31 | sean-k-mooney | oh | |
| 13:18:36 | sean-k-mooney | its an atribute you can set 9pfs | |
| 13:18:42 | Uggla_ | hum are you sure, I can configure it using virtual machine manager, and I think it uses libvirt. | |
| 13:21:19 | sean-k-mooney | it might be setting things qit qemu args | |
| 13:23:40 | Uggla_ | it looks like the conf is: https://paste.opendev.org/show/bC6FS8a7RP2J0tT9G2Jn/ | |
| 13:24:42 | sean-k-mooney | ya so it looks like its the defautl driver if you dont use virtio-fs | |
| 13:26:22 | Uggla_ | I have not dig into it, this is just the conf porposed by virtual machine manager. I'm a bit afraid about the <address type="pci"... as I don't really know the purpose so far. | |
| 13:26:29 | sean-k-mooney | im not sure that readonly is really required but perhaps we should consider it | |
| 13:26:58 | Uggla_ | sean-k-mooney, I have discussed and open a ticket to the virt team to have it supported. | |
| 13:27:03 | sean-k-mooney | Uggla_: thats just the adress in the vm where its mounted | |
| 13:27:18 | sean-k-mooney | Uggla_: we can rely on it however fo the intial enablement | |
| 13:27:39 | sean-k-mooney | we have a policy of rquiring features to be in a released version of libvirt/qemu before we can use it in nova | |
| 13:28:10 | sean-k-mooney | i woudl perhaps isnteead consier addign ahard limit on the virtiofs size and allwo it to be read write | |
| 13:28:18 | sean-k-mooney | but lets see what others think | |
| 13:33:58 | Uggla_ | one of the benefits is that virtio9p seems to not have issue with snapshot etc... but probably less performant than virtiofs. | |
| 13:35:00 | Uggla_ | I think virtio9p does not require shared memory that's the reason why. | |
| 13:38:17 | Uggla_ | sean-k-mooney, regarding virtiofs if we really want ro we can work around it using a bind mount and explain it it the doc. | |
| 13:39:33 | Uggla_ | sean-k-mooney, do you thing it worth to also write a spec for virtio9p ? | |
| 14:13:19 | sean-k-mooney | no | |
| 14:13:28 | sean-k-mooney | i dont think we shoudl add supprot for virtio9p | |
| 14:13:49 | sean-k-mooney | it has much lower performacne and it will likely get remvoed eventully once virtiofs has reach feature partiy | |
| 14:30:41 | opendevreview | Danylo Vodopianov proposed openstack/nova-specs master: Add support for Napatech LinkVirt SmartNICs https://review.opendev.org/c/openstack/nova-specs/+/859290 | |
| 14:53:38 | opendevreview | Danylo Vodopianov proposed openstack/nova-specs master: Add support for Napatech LinkVirt SmartNICs https://review.opendev.org/c/openstack/nova-specs/+/859290 | |
| 15:13:24 | opendevreview | Danylo Vodopianov proposed openstack/nova-specs master: Add support for Napatech LinkVirt SmartNICs https://review.opendev.org/c/openstack/nova-specs/+/859290 | |
| #openstack-nova - 2023-06-02 | |||
| 12:00:50 | opendevreview | Vasyl Saienko proposed openstack/os-vif master: Don't break traffic if port already exists https://review.opendev.org/c/openstack/os-vif/+/885127 | |
| 12:01:47 | opendevreview | Vasyl Saienko proposed openstack/os-vif master: Don't break traffic if port already exists https://review.opendev.org/c/openstack/os-vif/+/885127 | |
| 12:04:20 | gibi | sean-k-mooney: we lost your +2 from https://review.opendev.org/c/openstack/nova/+/873216 | |
| 12:04:29 | gibi | ahh sorry | |
| 12:04:38 | gibi | it wasn't review +2 it was prio +2 | |
| 12:04:40 | gibi | then never mind | |
| 12:04:52 | gibi | anyhow I reviewed the fix and I'm +2 on it | |
| 13:28:36 | sean-k-mooney | gibi: i can review it now if your +2 its probly mostly fine | |
| 13:29:37 | sean-k-mooney | actully i think i planned to +2 that previosuly yesterday or the day before | |
| 13:30:49 | sean-k-mooney | ya i had intended to add both CR: +2 and RP: +2 | |
| 15:10:23 | gibi | :) | |
| #openstack-nova - 2023-06-05 | |||
| 11:01:38 | mdbooth | I'll be running a forum session on Kubernetes on OpenStack in Vancouver next week. It's for users and developers of all related projects to talk to each other. Etherpad is here if there's anything you'd like to discuss: https://etherpad.opendev.org/p/openinfra-2023-kubernetes-on-openstack | |
| 11:57:22 | opendevreview | Amit Uniyal proposed openstack/nova master: Reproducer for dangling bdms https://review.opendev.org/c/openstack/nova/+/881457 | |
| 11:57:23 | opendevreview | Amit Uniyal proposed openstack/nova master: Delete dangling bdms https://review.opendev.org/c/openstack/nova/+/882284 | |
| 15:53:10 | opendevreview | Amit Uniyal proposed openstack/nova master: Reproducer for dangling bdms https://review.opendev.org/c/openstack/nova/+/881457 | |
| 15:53:10 | opendevreview | Amit Uniyal proposed openstack/nova master: Delete dangling bdms https://review.opendev.org/c/openstack/nova/+/882284 | |
| 16:23:36 | opendevreview | Sylvain Bauza proposed openstack/nova master: cpu: fix the privsep issue when offlining the cpu https://review.opendev.org/c/openstack/nova/+/885293 | |
| 16:23:49 | bauzas | sometimes I hate myself ^ | |
| 16:23:54 | bauzas | gibi: dansmith: ^ | |
| 16:24:04 | bauzas | we need to backport it to Antelope | |
| 16:24:15 | dansmith | ouch | |
| 16:24:27 | bauzas | yeah :( | |
| 16:24:36 | bauzas | and the functests don't verify it | |
| 16:25:07 | bauzas | maybe I should also create a relnote | |
| 16:25:34 | bauzas | so operators would see to use the next Antelope .z release if they want to use it | |
| 16:26:05 | dansmith | maybe you'll need to make a release before your demo so people can actually use it :D | |
| 16:28:39 | bauzas | yeah, I'll prepare a release | |
| 16:28:52 | bauzas | and I'll tell it loudly | |
| 16:28:58 | bauzas | as I'm very sad :( | |
| 16:30:31 | bauzas | also we maybe should have some test that verify that we don't forget a privsep decorator | |
| 16:30:52 | bauzas | like we do with objects or other stuff | |
| 16:31:35 | bauzas | but that's a bit difficult | |
| 16:47:21 | dansmith | yeah, not sure how to do that | |
| 17:42:45 | opendevreview | Sylvain Bauza proposed openstack/nova master: cpu: fix the privsep issue when offlining the cpu https://review.opendev.org/c/openstack/nova/+/885293 | |
| 17:43:00 | bauzas | dansmith: added a relnote for making it more public | |
| 17:43:25 | bauzas | and again, sorry for this bug :( | |
| 17:43:38 | dansmith | how could permissions cause a FileNotFound/ | |