What I found interesting in your bug vs the upstream one is that you
have a single node in your nodeset, that might help to simplify things.
# new guest with UVT (any other probably does as well)
$ uvt-kvm create --host-passthrough --password=ubuntu h-migr-nodeset
release=hirsute arch=amd64 label=daily
# verify things migrate
virsh migrate --unsafe --live h-migr-nodeset
qemu+ssh://testkvm-hirsute-to/system
# make it use HP+nodeset
Adding this scction to the config
<memoryBacking>
<hugepages>
<page size='2048' unit='KiB' nodeset='0'/>
</hugepages>
</memoryBacking>
This migrates just fine in Hirsute (qemu 5.1 / libvirt 6.9)
You mentioned to see this on Bionic, so I tried the same on Bionic - but that
works as well.
Hmm, so far trying to simplify this (e.g. take openstack out of the equation)
fails.
Once you get to create VMs manually could you check what the smallest set of
"ingredients" is to trigger the issue?
Note: this is a non Numa system, I only have node 0 - this could be
important. It would be awesome if we could make it trigger on those, but
if eventually a numa bare metal system is required we can't change it.
P.S: TBH - I'm not sure if that isn't just a real limitation, but let us track
the case until we know for sure. Or did you see that ever working and it
degraded in a newer release?
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1905424
Title:
Live migration failure
To manage notifications about this bug go to:
https://bugs.launchpad.net/libvirt/+bug/1905424/+subscriptions
--
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
