Description of problem:
A regression was introduced during the introduction of snapshot-based mirroring that results in the fast-diff feature being broken.
"rbd du" will not only take orders of magnitude longer, it will also potentially return incorrect results if the fast-diff feature is not enabled but object-map is enabled.
"rbd-mirror" in snapshot-based mirroring mode will not be able to use the speed optimization of only testing dirty blocks and instead will query deltas for each data block from remote OSDs. If the fast-diff feature is not enabled and object-map is enabled, it will result in data corruption.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. create a large image
2. 'rbd du' against the image
Results can take minutes for large enough images.
Should execute almost immediately because fast-diff will be used
Note: this might also effect the dashboard since it calculates RBD image disk usage in the background if fast-diff feature is enabled, but the librbd API is not correctly using fast-diff.
Hi Jason, From the description the bZ seems to be a regression and don't have workaround. We should include this BZ into 4.2
Hi Jason, Thanks for sharing the details. Proposing the BZ for 4.2 based on the potential impact to OSP customers.
Verified on ceph version 14.2.11-95.el8cp
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory (Important: Red Hat Ceph Storage 4.2 Security and Bug Fix update), and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.