just adding some more info.
i also deploy openstack rocky on a ubuntu 18.04 host
Linux cloud-5 4.15.0-43-generic #46-Ubuntu SMP Thu Dec 6 14:45:28 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
with ubuntu 18.04 l1 guest running
Linux numa-migration-1 4.15.0-45-generic #48-Ubuntu SMP Tue Jan 29 16:28:13 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
and when i lauch a cirros l2 guest i get very similar kvm_intel call traces.
if i upgrade the l1 guest kernel to the latest linux-virtual-hwe-18.04-edge kernel
which is currently
Linux numa-migration-2 4.18.0-15-generic #16~18.04.1-Ubuntu SMP Thu Feb 7 14:06:04 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
i still get the same crashes when i launch the l2 cirros image
i was originally wondering if this was some how related to the gust cpu model
so i tried seting the l1 and l2 guest cpu_mode to host-passthough
to maximies the feature that were availabel tot he guest
i then also tried to limit the l2 guest to the kvm64 cpu-model so that the l2 guest
used the minium set of kvm feature but this did not seam to have an benifical
effect.
the host is useign
Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz cpus which i have used in the past with
ubuntu 16.04 and nested virt without issue so this seams to be related to the 18.04 kenel in some way.
the fact that i am seeing the same issue on a ubuntu 18.04 host with an 18.04 l1 guest
and that gyurco is only seeing this with an 18.04 l1 guest makes me think that this is likely an issue with kvm in the l1 guests kernel not the hosts kernel.
im going to try and confrim that with a centos 7 l1 guest tommorow but since i would like to be
able to do python 3.6/3.7 testing in my local settup i would prefer to use a Ubuntu 18.04 l1 guest in the long run so i hope this can be resolved.
just adding some more info.
i also deploy openstack rocky on a ubuntu 18.04 host
Linux cloud-5 4.15.0-43-generic #46-Ubuntu SMP Thu Dec 6 14:45:28 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
with ubuntu 18.04 l1 guest running
Linux numa-migration-1 4.15.0-45-generic #48-Ubuntu SMP Tue Jan 29 16:28:13 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
and when i lauch a cirros l2 guest i get very similar kvm_intel call traces.
if i upgrade the l1 guest kernel to the latest linux-virtual- hwe-18. 04-edge kernel
which is currently
Linux numa-migration-2 4.18.0-15-generic #16~18.04.1-Ubuntu SMP Thu Feb 7 14:06:04 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
i still get the same crashes when i launch the l2 cirros image
i was originally wondering if this was some how related to the gust cpu model
so i tried seting the l1 and l2 guest cpu_mode to host-passthough
to maximies the feature that were availabel tot he guest
i then also tried to limit the l2 guest to the kvm64 cpu-model so that the l2 guest
used the minium set of kvm feature but this did not seam to have an benifical
effect.
the host is useign
Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz cpus which i have used in the past with
ubuntu 16.04 and nested virt without issue so this seams to be related to the 18.04 kenel in some way.
the fact that i am seeing the same issue on a ubuntu 18.04 host with an 18.04 l1 guest
and that gyurco is only seeing this with an 18.04 l1 guest makes me think that this is likely an issue with kvm in the l1 guests kernel not the hosts kernel.
im going to try and confrim that with a centos 7 l1 guest tommorow but since i would like to be
able to do python 3.6/3.7 testing in my local settup i would prefer to use a Ubuntu 18.04 l1 guest in the long run so i hope this can be resolved.