Bug 151815

Summary:	Backup/Restore are very slow on OCFS with 2.4.21-20.ELsmp
Product:	Red Hat Enterprise Linux 3	Reporter:	Ashish Tripathi <atripathi>
Component:	kernel	Assignee:	Larry Woodman <lwoodman>
Status:	CLOSED NOTABUG	QA Contact:	Brian Brock <bbrock>
Severity:	medium	Docs Contact:
Priority:	medium
Version:	3.0	CC:	coughlan, jbaron, petrides, riel
Target Milestone:	---
Target Release:	---
Hardware:	i386
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2006-12-08 13:48:44 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Ashish Tripathi 2005-03-22 16:48:50 UTC

From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1; .NET CLR 1.1.4322)

Description of problem:
Hi,
  I am using Redhat AS 3.0 (Update 3) with "2.4.21-20.ELsmp" Kernel with a 2 node RAC running OCFS (ocfs-2.4.21-EL-smp-1.0.13-1).Shared disks reside on a HP MA8000 SAN and are connected to the server using QLogic 2340 HBAs.Every server has only one HBA.Before upgrading to Kernel 2.4.21-20.ELsmp I was getting decent performace with tape backups and restores which is done over the network using HP Data Protector 5.but after the upgrade the performance of backups/restores have degraded considerabley.One intresting thing I noted was that when the backup is running and if I start a dd read process on the other shared file system in the same server, the backup speed picks up almost 2-3 times.But doing the same with restore doesn't help at all.
Is there any bug with kernel 2.4.21-20.ELsmp which is causing this behavior?
Thanks.
Ashish

Version-Release number of selected component (if applicable):
2.4.21-20.ELsmp

How reproducible:
Didn't try


Additional info:

Comment 1 Larry Woodman 2005-04-04 18:02:40 UTC

Kernel 2.4.21-20 was RHEL3-U3 and that kernel had a bug in which inodes were not
preperly reclaimed.  That bug was fixed in RHEL3-U4/kernel-2.4.21-27, can you
grab that kernel or even the latedt RHEL3-U5 kernel and see if the problem you
described is fixed?

Thank You, Larry Woodman

Comment 2 Ashish Tripathi 2005-04-13 17:05:20 UTC

Hi,
   I installed "2.4.21-27.0.2.ELsmp" kernel and ran the backups again but 
still the same result. By running a "dd read" process on another filesystem 
which is using a separate disks (LUN) makes the backup to run almost 10 times 
faster.
Thanks.
Ashish

Comment 3 Larry Woodman 2005-04-13 17:11:10 UTC

At this point I think we need to have Oracle look at this sice its OCFS and from
what I can tell its not happening on other file systems.

Larry

Comment 4 Ashish Tripathi 2005-04-14 20:34:11 UTC

Hi,
   I did a test by creating a EXT3 filesystem on the same disks.The backup 
speed was fine but the restore to the same disks using the same backup was 
half slower.
Can there be a bug in the drive of Qlogic card?

Thanks.
Ashish