On Wed, Jan 07, 2026 at 04:41:09PM +0530, Prasad Pandit wrote:
> On Tue, 6 Jan 2026 at 20:39, Peter Xu <[email protected]> wrote:
> > Copying Dan and Jiri in case ...
> 
> * Thank you for copying.
> 
> >>           } else if (s->current_state == ACTIVE && s->trigger == 
> >> 'error-occurred') {
> >>               s->current_state = STOP
> >>               s->reason = "Error occurred, migration failed"
> >
> > We can't change status that were already used, like FAILED.  Libvirt and
> > all mgmt may rely on it.
> 
> * True; If we decide to go in that direction, we'll have to tell
> libvirtd(8) and others about new states.
> 
> > Personally I don't see much benefit on adding a new "trigger" internal API.
> > If we want to forbid some state machine transitions, we can use a
> > transition map.  Said that, IMHO it's separate from what we're discussing
> > here.
> 
> * If we can reduce/rationalise the current 17-18 states and related
> complexity, it'll help to simplify things.

Personally I still don't see the benefit of doing so.

I agree MIGRATION_STATUS_CANCELLING isn't a must to have, but since we have
it already and libvirt should treat it almost the same as ACTIVE we don't
have problem with it either.  It doesn't sound like a big issue.

I don't see how we can shrink the rest of status otherwise, or what would
you propose for removal?

Thanks,

-- 
Peter Xu


Reply via email to