Bug #64016: rgw crashes can leave behind versioned bucket index entries that can't be cleaned up - rgw - Ceph

Actions

Copy link

Bug #64016

open

rgw crashes can leave behind versioned bucket index entries that can't be cleaned up

Added by Cory Snyder 4 months ago. Updated 4 months ago.

Status:

Fix Under Review

Priority:

Normal

Assignee:

Cory Snyder

Target version:

% Done:

Source:

Community (dev)

Tags:

Backport:

Regression:

Severity:

2 - major

Reviewed:

Affected Versions:

ceph-qa-suite:

Pull request ID:

55165

Crash signature (v1):

Crash signature (v2):

Description

When an RGW daemon crashes or is killed in a manner that doesn't allow pending OSD ops to be flushed through the network, it can leave behind index entries in an inconsistent state. These index entries cannot be cleaned up by existing radosgw-admin commands or by subsequent deletes. These entries can also cause the bucket object count to be inaccurate in a manner that is not fixable by radosgw-admin commands.