Bug 1066167 - GlusterFS Snapshot delete of attached volume fails if it runs > 10 minutes
Summary: GlusterFS Snapshot delete of attached volume fails if it runs > 10 minutes
Keywords:
Status: CLOSED UPSTREAM
Alias: None
Product: Red Hat OpenStack
Classification: Red Hat
Component: openstack-nova
Version: 4.0
Hardware: All
OS: All
urgent
urgent
Target Milestone: ---
: 6.0 (Juno)
Assignee: Eric Harney
QA Contact: nlevinki
URL:
Whiteboard: storage
Depends On:
Blocks: 1045196 1056037
TreeView+ depends on / blocked
 
Reported: 2014-02-17 22:03 UTC by Eric Harney
Modified: 2019-09-09 15:55 UTC (History)
8 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Nova code change to correspond to Cinder fix in bug 1056037.
Clone Of: 1056037
Environment:
Last Closed: 2014-10-09 13:26:01 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Launchpad 1273984 0 None None None Never
OpenStack gerrit 69761 0 None None None Never

Description Eric Harney 2014-02-17 22:03:02 UTC
+++ This bug was initially created as a clone of Bug #1056037 +++

...snip...

--- Additional comment from Eric Harney on 2014-01-21 13:57:22 EST ---

The basic problem here is that Cinder has a fixed time out when waiting for snapshot_delete operations on the Nova side to complete.  If they take too long (even when things are functioning correctly) Cinder will prematurely fail the operation.

To fix this, we need to have Nova send back updates of job percent complete while the block job is in-progress.  Cinder can then reset its timeout window based on these updates.  (This should be doable without changing how the APIs work between Cinder and Nova today.)

Comment 2 Eric Harney 2014-03-10 14:14:41 UTC
Moving along with bug 1056037.

Comment 3 Ben England 2014-04-07 19:07:53 UTC
why does it take so long to delete a snapshot?  Does it have to scrub the snapshot file before deleting it for security reasons?

Comment 4 Eric Harney 2014-04-07 19:28:01 UTC
(In reply to Ben England from comment #3)

We are forced to do a blockRebase rather than a commit when deleting the only remaining snapshot, which means the full volume data must be copied.  This is a libvirt limitation.

Comment 6 Sean Cohen 2014-10-09 13:26:01 UTC
This likely indicates using a version of libvirt which had known bugs in it in this area. Closing pending further info on reproduction.


Note You need to log in before you can comment on or make changes to this bug.