Open Stack

On Wed, Mar 13, 2013 at 3:58 PM, Lorin Hochstein
<lorin at nimbisservices.com> wrote:
> Michael:
>
> Is this related to the remove_unused_base_images flag?
>
> cf
> https://review.openstack.org/#/c/22040/
> https://bugs.launchpad.net/nova/+bug/1126375
>
> If so, I think this is a strong argument for changing the default.

Yes, this is the flag we're talking about. However, the bug was fixed
when it was reported by Nectar quite some time ago for grizzly.
Unfortunately, this operator ran his own shell script, which behaved
poorly with shared storage. This I suspect is because there is a lot
of advice out on the web which dates back to essex, where this problem
was handled manually.

The fix is in commit c2de33a0a2132774dc295861cef138ec24bb0cf9 for
grizzly, which was not back ported to folsom (see
https://review.openstack.org/#/c/16726/ for details). However, the
image cache manager is now off by default in folsom (its interval is
set to zero, which means don't run).

My comments from the code review you link to still hold -- there is
more danger in an unbounded disk cache then there is in shared NFS
storage, especially now that we've realised that edge condition exists
and handled it. Please note that this bug is only present for users
using shared NFS storage for instances, which doesn't seem to be
particularly common.

Michael