Comment 26 for bug 1882521

Revision history for this message
Lee Yarwood (lyarwood) wrote : Re: Failing device detachments on Focal

Continuing to see this again after the break, I've had a look at some example libvirtd logs an unlike https://bugs.launchpad.net/qemu/+bug/1894804 we don't even see the first DEVICE_DELETED event being emitted from QEMU now.

This could still be the same issue *if* n-cpu is asking libvirtd to send the second device_del command *before* QEMU has had change to emit the first DEVICE_DELETED event. So we could again extend the time between requests from n-cpu to libvirtd but that's another hackaround.

I've highlighted again to gibi that the only real fix within Nova for this is to move to a libvirtd event based approach as outlined during the PTG and I had started to PoC below:

https://review.opendev.org/c/openstack/nova/+/749929