On 05.02.26 17:51, Chaney, Ben wrote:
On 2/5/26, 3:12 AM, "Vladimir Sementsov-Ogievskiy" <[email protected]
<mailto:[email protected]>> wrote:
On 04.02.26 19:46, Chaney, Ben wrote:
Is it possible to relax this constraint at all? If so, I
would prefer to allow script= and downscript= parameters
to remain in place.
That possible, but this requires some additional logic I think.
What if migration fails? Who should call downscript? Migration
may be successful on source, and fail on target.. In this case,
management tool usually resume stopped source. And in this case
source should get again a responsibility to call downscript.
Hmm... If script and downscript are not set, they default to
/etc/qemu-ifup and /etc/qemu-ifdown. To disable them
altogether you must pass script=no,downscript=no, which
is not currently possible with this patch.
Yes, but when new option is set, we don't use any scripts. Like
for other options which doesn't allow use of script/downscript
(fd=, fds=, helper=).
But I see now, it was wrong decision.
So, I think it should be additional patch on top, which introduce
support for script/downscript together with backend-transfer.
Would it be possible to at least support script=no,downscript=no as
part of this patch? We may want to require it because it sounds like
we don't have the logic to call qemu-ifup and qemu-ifdown correctly
in the event of a failure.
Agree, that's correct. So, for this patch I should require script
and downscript to be present and set to "no", to be able make a
full support for them later.
--
Best regards,
Vladimir
