Bug #15378: ceph_volume_client: hasty removal of OSD caps during deauthorization - CephFS - Ceph

Actions

Copy link

Bug #15378

closed

ceph_volume_client: hasty removal of OSD caps during deauthorization

Added by Ramana Raja about 8 years ago. Updated about 5 years ago.

Status:

Rejected

Priority:

Normal

Assignee:

Ramana Raja

Category:

Target version:

% Done:

Source:

Development

Tags:

Backport:

Regression:

Severity:

3 - minor

Reviewed:

Affected Versions:

ceph-qa-suite:

Component(FS):

Labels (FS):

Manila

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

When deauthorizing access of a Ceph user to a volume (a CephFS directory),
the volume client removes OSD caps that allow access to the data pool
used by the volume. The deauthorization method does not consider that the
user would still need access to the data pool if it had already been
authorized to access other volumes using the same pool.

Actions

Copy link

Updated by Greg Farnum about 8 years ago

Category changed from 44 to 87
Assignee set to Ramana Raja

Actions

Copy link

Updated by Ramana Raja about 8 years ago

This issue would no longer be a concern once each volume (CephFS directory) gets its own RADOS namespace http://tracker.ceph.com/issues/15400
So when a auth ID is authorized mount access to a volume, the auth ID's OSD caps would be restricted to a RADOS namespace (unique to a volume) and not to an entire data pool. Likewise, the deauthorization of volume access for user, would involve removing OSD caps for access to the RADOS namespace that is specific to the volume.

Actions

Copy link