Cancellation of VM core dumps

[Jiri Denemark]

On Tue, Oct 05, 2021 at 14:30:43 +0000, Simon Rowe wrote:
> I see that cancellation of core dump jobs that are memory-only are explicitly rejected
> 
> 
> >>> d.abortJob()
> 
> libvirt: QEMU Driver error : Requested operation is not valid: cannot abort memory-only dump
> 
> Traceback (most recent call last):
> 
>   File "<stdin>", line 1, in <module>
> 
>   File "/usr/lib64/python2.7/site-packages/libvirt.py", line 544, in abortJob
> 
>     if ret == -1: raise libvirtError ('virDomainAbortJob() failed', dom=self)
> 
> libvirt.libvirtError: Requested operation is not valid: cannot abort memory-only dump
> 
> Is there a technical reason for this? Dumps of very large guests may take a long time and it would be useful to be able to abort them.

Yes. Historically, dumps (not the memory-only ones) were implemented
using migration to a file and that process can easily be aborted.
However, memory-only dumps use a special QEMU command and thus can't be
aborted in the same way libvirt abort migrations. That said we would
need to modify libvirt and likely QEMU as well to be able to abort
memory-only dumps.

Jirka