opendevreview | Ilia Baikov proposed openstack/nova master: api: Add cloud-init as whitelisted property in DynamicJSON vendor data to not be JSON-encoded https://review.opendev.org/c/openstack/nova/+/917276 | 00:42 |
---|---|---|
songwenping__ | sean-k-mooney: how can i get the access to the bugzilla? https://bugzilla.redhat.com/show_bug.cgi?id=953339 | 01:46 |
sean-k-mooney | songwenping_: its restircted to redhat employee's its also for rhel 6.4 which has been out of supprot for seveeral years so im not sure its going to be relvent ot you? | 10:50 |
sean-k-mooney | it was fixed in 2014 | 10:51 |
songwenping_ | sean-k-mooney: we meet the same problem with the libvirt 6.2.0, is the libvirt is too old and which version i will upgrade? | 10:57 |
sean-k-mooney | not being able to attach >50 VF to a single vm? | 10:58 |
sean-k-mooney | the bug you linked was in libnl which is the netlink libary but hte libvirt fix that used it was in 0.10.2 | 10:59 |
sean-k-mooney | so this is not because your libvirt is two old | 11:00 |
sean-k-mooney | actully it might not require 50 VF with one vm just vfs 50 and above? | 11:01 |
sean-k-mooney | its really not clear form the bug | 11:02 |
songwenping_ | only bind one VF to vm | 11:02 |
songwenping_ | the full error message from nova-compute is : Instance failed to spawn: libvirt.libvirtError: internal error: missing IFLA_VF_INFO in netlink response | 11:02 |
sean-k-mooney | songwenping_: ok my guess is there my be a similar bug but you are not hitting that one | 11:03 |
songwenping_ | so which version of netlink libary is support? | 11:05 |
sean-k-mooney | that has noting to do with nova | 11:05 |
sean-k-mooney | that is determined by your distro | 11:05 |
sean-k-mooney | what card are you tring to use | 11:05 |
sean-k-mooney | i found this googleing https://forums.developer.nvidia.com/t/connectx-6-dx-sr-iov-libvirt-error/249762 | 11:06 |
songwenping_ | we are using Mellanox cx5 nic | 11:08 |
sean-k-mooney | it shared the same kernel driver so it may be the same or a simialr bug | 11:09 |
songwenping_ | yeah, i alse found the site, so i wonder if we should upgrade the libvirt | 11:10 |
elodilles | bauzas: if you are around, a gentle reminder for the nova release patches o:) https://review.opendev.org/q/project:openstack/releases+file:%5Edeliverables/.*/nova.yaml+is:open | 12:04 |
elodilles | Uggla_ auniyal : fyi ^^^ | 12:05 |
frickler | oh, that's a nice search filter, /me takes notes ;) | 12:09 |
bauzas | frickler: yeah I used the filter something like 9yo when I was only a nova-scheduler expert :) | 12:11 |
samcat116 | Hi all, I submitted this bug about Placement and aarch64 nodes receiving x86 traits about a week ago. Any ideas on a next step I could take? https://bugs.launchpad.net/nova/+bug/2062425 Thanks! | 15:13 |
tkajinam | samcat116, If I read the current code correctly you still have to add HW_ARCH_* traits manually to resource providers. I don't see any logic in libvirt driver to report these traits. | 15:39 |
tkajinam | (I've left a comment in a bug too | 15:39 |
samcat116 | I see all of the relevant x86 traits created on the other x86 nodes, and I did nothing to create those | 15:40 |
tkajinam | I was not involved in the discussion of that cpu arch support. probably someone from the core can double-check my observation and probably give context of the current implementation | 15:40 |
tkajinam | that's a trait for CPU feature, not one for arch | 15:40 |
tkajinam | no cpu archutecture trait is currently reported automatically. | 15:41 |
samcat116 | Oh so those are different things | 15:42 |
tkajinam | cpu architecture trait for x86_64 would be HW_ARCH_X86_64 | 15:42 |
tkajinam | samcat116, yeah that's different | 15:44 |
tkajinam | it's ugly that HW_CPU_X86_AESNI is reported in aarh64 node, but we first have to have a mechanism to detect the cpu architecture | 15:44 |
tkajinam | which is not implemented now | 15:44 |
samcat116 | Ok, so I guess this is more of a feature request then to automatically set that | 15:44 |
samcat116 | Yeah I mean I see on startup nova-compute recognizes its aarch64 and tells me its not really tested | 15:45 |
samcat116 | But from what I've read the support should be pretty decent by now | 15:45 |
tkajinam | yeah, I agree > more of a feature request | 15:45 |
samcat116 | am I also correct in understanding that if I manually apply the right arch trait to this node, it will lose all of the automatic traits it has now? | 15:45 |
tkajinam | iirc if you use `openstack resource provider trait set` to add trait you have to give all traits | 15:48 |
tkajinam | once additional traits are added, these should be kept. | 15:50 |
samcat116 | Ok, I'll try that and manually add all of the ones it got automatically | 15:52 |
samcat116 | thank you! | 15:52 |
opendevreview | Pierre Riteau proposed openstack/nova master: Fix formatting issues in extra-specs docs https://review.opendev.org/c/openstack/nova/+/917462 | 17:07 |
opendevreview | Amit Uniyal proposed openstack/nova master: Note the deleyad address view https://review.opendev.org/c/openstack/nova/+/827856 | 17:31 |
sean-k-mooney | bauzas: fyi i created https://blueprints.launchpad.net/nova/+spec/eventlet-removal-part-1 | 18:50 |
opendevreview | Merged openstack/nova master: Note the deleyad address view https://review.opendev.org/c/openstack/nova/+/827856 | 18:59 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!