[openstack-dev] Discussion of Nova Bug 1189543 (Setting drive error policy in Nova)

Robert Collins robertc at robertcollins.net
Mon Jun 10 19:07:18 UTC 2013


On 11 June 2013 06:38, Solly Ross <sross at redhat.com> wrote:
> I think the idea is that a VM's OS might just blindly go on writing, or even stop the write and not write some important piece of data.  The idea behind pausing the VM would be that someone could hypothetically fix the error (allocate more space for the disk, etc) and resume the VM, thus not losing the data that was to be written.

But that applies to bare metal environments too, no? If your DB has a
write error, it needs to know about it - in fact, unless qemu will
buffer and re-try the write itself when the issue is fixed, pausing
the VM is *more* likely to lead to data loss, because the VM won't be
told, and the write still failed. Unless I misunderstand something?

Anyhow, seems like a super-dangerous thing to do, very much surprising
to folk who 'own' the machines.

-Rob



-- 
Robert Collins <rbtcollins at hp.com>
Distinguished Technologist
HP Cloud Services



More information about the OpenStack-dev mailing list