Bug 1187821

Summary:	.rgw pool contains extra objects
Product:	[Red Hat Storage] Red Hat Ceph Storage	Reporter:	tbrekke
Component:	RGW	Assignee:	Yehuda Sadeh <yehuda>
Status:	CLOSED ERRATA	QA Contact:	ceph-qe-bugs <ceph-qe-bugs>
Severity:	medium	Docs Contact:
Priority:	unspecified
Version:	1.2.2	CC:	cbodley, ceph-eng-bugs, flucifre, hnallurv, jherrman, kbader, kdreyer, mbenjamin, nlevine, owasserm, smanjara, sweil, yehuda
Target Milestone:	pre-dev-freeze
Target Release:	1.3.0
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:	ceph-0.94.1-3.el7cp	Doc Type:	Bug Fix
Doc Text:	Recreating a previously existing node in RGW previously did not remove the node instance metadata object and thus created a redundant object in the RGW pool. This update addresses this problem, and redundant objects are no longer generated in the described scenario.	Story Points:	---
Clone Of:
Clones:	1212524 (view as bug list)		Environment:
Last Closed:	2015-06-24 15:50:12 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:
Bug Blocks:	1212524

Description tbrekke 2015-01-30 22:07:56 UTC

Description of problem:

Recreating a bucket via s3 created a new object in the .rgw pool. 

Steps to Reproduce:
1. Create a bucket with n s3 client
2. Delete bucket, then recreate the bucket
3. radosgw-admin metadata get bucket:<bucket> shows the correct bucket_id
4. `rados -p .rgw ls` shows the extra object that do not match the bucket_id


Expected results:
Deleting a bucket should remove the bucket metadata. 

Additional info:
While this is typically not large issue, it is possible for millions of objects to end up in the .rgw pool.

Comment 1 Federico Lucifredi 2015-03-25 23:55:31 UTC

Yehuda, what say you?

Comment 2 Federico Lucifredi 2015-03-25 23:57:04 UTC

Yehuda, what say you? 

If you are +1 on this, please set devel flag accordingly, thanks.

Comment 3 Yehuda Sadeh 2015-03-26 00:09:58 UTC

Updated the upstream bug number. Orit is working on this specific bug. The reason we keep the old object metadata around is that in the case of multi zone configuration we don't want to lose the knowledge about an old bucket instance, in the case it is still referenced by another zone, and we have no way to know whether it's needed or not.

Comment 4 Yehuda Sadeh 2015-04-07 17:45:36 UTC

Fix was merged upstream to commit:dfdc7afb59cc8e32cf8bff55faa09076c853de06.

Comment 5 Ken Dreyer (Red Hat) 2015-04-16 14:51:53 UTC

I'll clone this bug so we can track the fix separately in RHCS 1.2 (based on Firefly, upstream's v0.80.8) and RHCS 1.3 (based on Hammer, upstream's v0.94.1)

Comment 6 Ken Dreyer (Red Hat) 2015-04-16 15:09:59 UTC

I've verified that dfdc7afb59cc8e32cf8bff55faa09076c853de06 cherry-picks cleanly onto hammer.

Comment 10 Ken Dreyer (Red Hat) 2015-04-16 23:19:36 UTC

It looks like fe158ecc25feefcea8aea4133118e4a84900a8ec upstream doesn't cherry-pick cleanly to 0.94.1. Yehuda, would you mind providing a cherry-pick for 0.94.1?

Comment 13 shilpa 2015-06-15 08:15:48 UTC

Verified on ceph-0.94.1-11.el7cp.x86_64. Deleting bucket also removes its metadata.

Comment 15 errata-xmlrpc 2015-06-24 15:50:12 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, 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-2015:1183