On Sat, Feb 17, 2007 at 01:22:35PM +0000, Mark McLoughlin wrote:
On Fri, 2007-02-16 at 19:08 +0000, Daniel P. Berrange wrote:
> On Fri, Feb 16, 2007 at 02:44:46PM +0000, Mark McLoughlin wrote:
> > Handle SIGHUP by shutting down all guests and networks
> > and re-loading configs
>
> This violates the 'principle of least surprise'. I certainly do not expect
> a config file reload to terminate all active guests. Since QEMU is a full
> virt system, there is no graceful shutdown process, and so this is equivalent
> to ripping the power cable out of all your VMs. IMHO this makes SIGHUP
> essentially useless.
>
> We already have the ability to store a secondary config against any active
> guest VMs, which will be automatically activated upon next boot of the
> guest - this is how we let virDefineDomain() override the config of an
> existing VM. IMHO, sending SIGHUP to the daemon should just scan for
> config files and if any VMs are active, load into the secondary config.
All good points, fixed now with the attached patch.
Looks good - simpler than I expected it to be - I forgot that the
qemudScanConfigs function would already do the right thing if being
run over.
> > Handle SIGTERM/SIGINT by cleanly shutting down
>
> Ack, this is desirable. We should also hook in SIGQUIT
Yep, done.
Dan.
--
|=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=|
|=- Perl modules:
http://search.cpan.org/~danberr/ -=|
|=- Projects:
http://freshmeat.net/~danielpb/ -=|
|=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|