slow OES restore

Hi,

I have a backup/restore application which uses SMS apis to do backup and restore. Restore is very slow when restoring to nss cluster volume. I debugged and found that, maximum time is consumed while writing the first chunk of data to the file. I doubt the slowness is during the file creation. To be exact, NWSMTSWriteDataSet call takes as long as 3 minutes and sometimes 6 minutes for completion. This happens only while writing the first buffer. For the rest of the buffers the operation completes very quick. SMS logs does not show any errors during the timeframe where slowness is seen.

Is there any parameters i can tune to overcome this slowness.

This is on OES2 cluster setup.
Compression is disabled
No Anti-Virus
Restore using SIDFparser for same set of files happens quick
  • On 08/03/2013 12:36, arunprabu wrote:

    > I have a backup/restore application which uses SMS apis to do backup
    > and restore. Restore is very slow when restoring to nss cluster volume.
    > I debugged and found that, maximum time is consumed while writing the
    > first chunk of data to the file. I doubt the slowness is during the file
    > creation. To be exact, NWSMTSWriteDataSet call takes as long as 3
    > minutes and sometimes 6 minutes for completion. This happens only while
    > writing the first buffer. For the rest of the buffers the operation
    > completes very quick. SMS logs does not show any errors during the
    > timeframe where slowness is seen.
    >
    > Is there any parameters i can tune to overcome this slowness.
    >
    > This is on OES2 cluster setup.
    > Compression is disabled
    > No Anti-Virus
    > Restore using SIDFparser for same set of files happens quick


    Which version(s) of OES2? (SP3 is latest supported, on SLES10 SP4)
    What patch level?
    Which backup application?

    HTH.
    --
    Simon
    Novell/SUSE/NetIQ Knowledge Partner

    ------------------------------------------------------------------------
    Do you work with Novell technologies at a university, college or school?
    If so, your campus could benefit from joining the Novell Technology
    Transfer Partner (TTP) program. See novell.com/ttp for more details.
    ------------------------------------------------------------------------
  • OES2 SP1
    kernel version 2.6.16.60-0.33

    Backup application is our own which uses sms apis. Backup/Restore runs fine on all standalone machines. This issue is seen only on one particular cluster setup.
  • When NWSMTSWriteDataSet is called, it takes 3 to 6 minutes to completion. The time taken is always in multiples of 30 seconds. I doubt if there is any configured timeout value which i am not sure. I monitored the sms logs when restore happens and the slowness happens exactly at the point when FS_ConvertNameToID is called.

    b11ffba0:GetObjectIDFromFid : Start
    b11ffba0:FS_ConvertNameToID : Start

    After this log there is a 4 minute delay and then the function returns with error code

    b11ffba0:FS_ConvertNameToID = fffffda7
    b11ffba0:GetObjectIDFromFid = fffffda7

    Can someone help me interpret these errors
  • In article <arunprabu.5rvxfz@no-mx.forums.novell.com>, Arunprabu wrote:
    > OES2 SP1
    >

    perhaps updating to a supported version (OES2 sp3, or OES11 sp1) would
    help you along.


    Andy Konecny
    Knowledge Partner (voluntary SysOp)
    KonecnyConsulting.ca in Toronto
    ----------------------------------------------------------------------
    Andy's Profile: http://forums.novell.com/member.php?userid=75037


  • > perhaps updating to a supported version (OES2 sp3, or OES11 sp1) would
    > help you along.
    This problem is seen on one of our customers production setup. They are not ready to upgrade.
  • peterkuo;2255540 wrote:
    You mean its OK on another cluster setup?

    The issue is seen in another cluster setup as well. I was wrong in the OP. Customer informed, it was seen in standalone system as well. There are 2 edirectories. Backup was done in a machine in first eDirectory and restored back to machines in second edirectory. I doubt the username who is owner of the restored files is not present in second eDirectory. That explains the following error
    F5C0CB70:GetFieldToProcessSection [4138 ]:fid = OWNER_NAME, dataSize=24
    f5c0cb70:GetObjectIDFromFid : Start
    f5c0cb70:FS_ConvertNameToID : Start
    f5c0cb70:FS_ConvertNameToID = fffffda7
    f5c0cb70:GetObjectIDFromFid = fffffda7

    But, i am not sure why the slowness is caused when this function is called
  • It has been a super long time since I developed anything using the SMS APIs but my experience with the NWSMTSWriteDataSet call is that it was fast - but that was as an NLM and not on Linux platform.

    Unfortunately I don't have the SMS library on my laptop so I can't even look up the docs. Will have to dig through some backup CDs ...