Bug 2021601 - rgw: deleting and purging a bucket can get stuck in an endless loop 4.2
Summary: rgw: deleting and purging a bucket can get stuck in an endless loop 4.2
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Red Hat Ceph Storage
Classification: Red Hat Storage
Component: RGW
Version: 4.2
Hardware: Unspecified
OS: Unspecified
unspecified
high
Target Milestone: ---
: 4.2z4
Assignee: J. Eric Ivancich
QA Contact: Tejas
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2021-11-09 17:45 UTC by J. Eric Ivancich
Modified: 2021-11-18 16:06 UTC (History)
10 users (show)

Fixed In Version: ceph-14.2.11-207.el8cp, ceph-14.2.11-207.el7cp
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed: 2021-11-18 16:06:02 UTC
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Issue Tracker RHCEPH-2208 0 None None None 2021-11-09 18:00:12 UTC
Red Hat Product Errata RHBA-2021:4740 0 None None None 2021-11-18 16:06:22 UTC

Description J. Eric Ivancich 2021-11-09 17:45:29 UTC
This bug was initially created as a copy of Bug #2021600

I am copying this bug because: 

It appears in multiple downstream releases.

Description of problem:

This stems from upstream tracker: https://tracker.ceph.com/issues/52662 .

If all unfinished multipart upload ".meta" index entries cannot be retrieved in one go, the marker is not advanced and the listing will be stuck. This happens during the phase of bucket deletion/purge where it tries to abort all incomplete multipart uploads.


Version-Release number of selected component (if applicable):


How reproducible:

Very reproducible.


Steps to Reproduce:
1. Load up the bucket index with many thousands of entries in the _multipart_ namespace representing parts of incomplete multipart uploads. Add two entries that are in that namespace spread apart in the lexically (so there are thousands of entries in between) and end with ".meta" each representing one incomplete multipart upload.
2. Issue `radosgw-admin bucket rm --bucket=<bucket> --purge-objects --bypass-gc`

Actual results:

Command will never finished. If --debug-rgw=20 added, one will see the same bucket index entries re-visited multiple times in the log. Bucket will not be removed.

Expected results:

Command to complete, bucket removed.

Additional info:

Comment 1 J. Eric Ivancich 2021-11-09 17:46:18 UTC
There is not an upstream nautilus PR. But there is an upstream octopus pr: https://github.com/ceph/ceph/pull/43863 .

Comment 4 J. Eric Ivancich 2021-11-10 15:55:28 UTC
Cherry-picked to add commit 9e2f28bd0efb267d657ffdcec35e5ea052f61da0 to ceph-4.2-rhel-patches.

Comment 14 errata-xmlrpc 2021-11-18 16:06:02 UTC
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 (Red Hat Ceph Storage 4.2 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.

https://access.redhat.com/errata/RHBA-2021:4740


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