[HPADM] fsadm defrag doesn't work as well as restoring files after newfs?

From: Jeff Lightner (jlightner_at_water.com)
Date: 11/02/05

  • Next message: Vidal, Ignacio - (Arg): "[HPADM] Use of gdb"
    Date: Wed, 2 Nov 2005 14:25:00 -0500
    To: <hpux-admin@DutchWorks.nl>
    
    

    We have a couple of filesystem that had 10 x 512 MB Oracle redo logs on
    each. We added space to the filesystems and the DBAs then expanded the
    10 redo logs from 512 MB to 1 GB each. They are now complaining about
    performance of the redo logs and claim it has to do with filesystem
    fragmentation.

    They claim the last time they did something similar (256 MB to 512 MB)
    that they had the admin copy the files out (with the DB down of course)
    then newfs the filesystem and copy the files back in to get rid of the
    fragmentation.

    Since we have OnlineJFS this seemed unnecessary to me and sure enough on
    reading and perusing forums found info on doing fsadm -ed for defrag.
    (-e for extent level and -d for directory level).

    As a test I used a separate filesystem and did prealloc to create 10 x
    512 MB files. I then did a remove of the first file and did prealloc to
    1 GB then a remove of the second file and a prealloc to 1 GB etc...
    until all 10 had been done. This all to simulate the fragmentation.

    Running ncheck and sorting/grepping for different information I was able
    to see the different extent ranges in use for each of the files. I also
    ran fsadm -ED to see the extent/directory fragmentation. On doing the
    fsadm -ed though I saw some but not a lot of improvement in
    fragmentation. On read of the man page I saw comments about maximum
    extent size allowed for move being modified with -l. Since default was
    64 and max was 2048 I tried the fsadm -ed -l 2048 to see if it helped
    but it didn't.

    However when I did the copy out to another filesystem then did a newfs
    on the original and copied back in sure enough ncheck showed a much
    better defragmentation. It appears therefore the fsadm defrag isn't of
    much use (at least for my 1 GB files).

    So am I missing something or is it just that the fsadm defrag really has
    no value for this purpose?

    Note: Please don't suggest "-F vxfs" as vxfs is in /etc/default/fs and
    per the man page that is the fstype it assume if -F isn't specified.

    Jeffrey C. Lightner
    Unix Systems Administrator
    DS Waters of North America
    678-486-3516

    --
                 ---> Please post QUESTIONS and SUMMARIES only!! <---
            To subscribe/unsubscribe to this list, contact majordomo@dutchworks.nl
           Name: hpux-admin@dutchworks.nl     Owner: owner-hpux-admin@dutchworks.nl
     
     Archives:  ftp.dutchworks.nl:/pub/digests/hpux-admin       (FTP, browse only)
                http://www.dutchworks.nl/htbin/hpsysadmin   (Web, browse & search)
    

  • Next message: Vidal, Ignacio - (Arg): "[HPADM] Use of gdb"

    Relevant Pages

    • Re: [PATCH] Avoiding fragmentation through different allocator
      ... James and Grant added to CC. ... > fragmentation for some orders are over time. ... > only support pagesize and reiser will create a filesystem with a blocksize ... > the assumption that high-order allocations will fail. ...
      (Linux-Kernel)
    • Re: fragmentation
      ... by fragmentation - because we use it in many different ways. ... the storage and computational overhead in maintaining the filesystem. ... have 8k or 16k allocation units, then even small files would have ... logically contiguous end up being discontiguous on the disk. ...
      (freebsd-questions)
    • Re: using star to move 3 millions files .. going slowly .. painfully
      ... at this poit I can wait the next 20 hours for star 1.4.2 to complete ... >> the fragmentation size on the filesystem is 2048 bytes then a lot of the ... >>Perhaps that is a feature of the Linux ext2fs filesystem. ... performance versus stripe depth would be almost sinusoidal within the region ...
      (comp.unix.solaris)
    • Re: Intel(r) Core?2 Duo Processors"
      ... Does top show a high load, or indicate that the system is ... suffer much from fragmentation, ... because he noticed that his computer sped up when given a fresh filesystem. ... You could try backing up and restoring to a fresh filesystem. ...
      (Fedora)
    • Re: Defragging drive possible in IRIX?
      ... job to defrag the drives at night, I don't know if it still is a default ... An XFS filesystem will only fragment if you are both low on RAM and low on disk ... temporary files (the worst for fragmentation), ...
      (comp.sys.sgi.admin)