SUMMARY: Ufsdump Backup Problem on Mirrored Soft Partition on V48 0

From: Gilliam, Kirk R. (KGilliam_at_bcbsm.com)
Date: 05/23/03

  • Next message: Homan, Charles (NE): "SUMMARY: (NO SUMMARY) Setting serial line receive buffer size?" 
    To: "'sunmanagers@sunmanagers.org'" <sunmanagers@sunmanagers.org>
Date: Fri, 23 May 2003 12:40:05 -0400

    Here is the original message:

    > Hello--
    >
    > I have a Sunfire 480R running Solaris 8 2/02 (s28s_u7wos_08a),
    > Generic_108528-18, with 4GB of ram and 2 - 900 MHz US-III+ CPU's and two
    > SEAGATE 36.42GB mirrored disks with soft partitions and the OBP level is
    > 4.6.4. The problem I am getting is my ufsdump backup fails for the slice
    > "/dev/md/rdsk/d91" with the following errors and doing a metastat on "d91"
    > reveals no errors. The 2 - 36GB disks are partitioned exactly alike. I
    > have no "disk failure" messages anywhere on the system. I have ran a
    > similar ufsdump on the two submirrors of the soft partition and I get the
    > same results, so I have no way of figuring out if 1 disk is bad over the
    > other or if somehow the soft partitioning is bad. Has anyone ever
    > encountered this before? I have the identical configuration running on a
    > Sunfire V120 (development server) with no problems.
    >
    > Please see output below. I will summarize.
    >
    > Thanks,
    >
    > Kirk Gilliam
    > Unix Systems Administrator
    > Blue Cross Blue Shield of Michigan
    >
    >
    > #######################################################################
    > Backup Error
    > #######################################################################
    >
    > 030406.232416 : Backing up /dev/md/rdsk/d91
    > DUMP: Writing 63 Kilobyte records
    > DUMP: Date of this level 0 dump: Sun Apr 06 23:24:16 2003
    > DUMP: Date of last level 0 dump: the epoch
    > DUMP: Dumping /dev/md/rdsk/d91 (psx401a2370:/app/sunone/meta) to
    > /dev/rmt/0n.
    > DUMP: Mapping (Pass I) [regular files]
    > DUMP: Mapping (Pass II) [directories]
    > DUMP: Estimated 817762 blocks (399.30MB).
    > DUMP: Dumping (Pass III) [directories]
    > DUMP: Dumping (Pass IV) [regular files]
    > DUMP: Warning - cannot read sector 6289353 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289354 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289355 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289356 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289357 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289358 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289359 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289392 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289393 of `/dev/md/rdsk/d91'
    > DUMP: Warning - cannot read sector 6289394 of `/dev/md/rdsk/d91'
    > .
    > .
    > .
    > DUMP: Warning - cannot read sector 6289455 of `/dev/md/rdsk/d91'
    > DUMP: More than 32 block read errors from dump device `/dev/md/rdsk/d91'
    > DUMP: NEEDS ATTENTION: Do you want to attempt to continue? ("yes" or "no")
    > DUMP: The ENTIRE dump is
    > aborted.
    > 030413.230342 : error during ufsdump
    >
    > #############################################################
    > Metastat Info
    > #############################################################
    >
    > d91: Mirror
    > Submirror 0: d71
    > State: Okay
    > Submirror 1: d81
    > State: Okay
    > Pass: 1
    > Read option: roundrobin (default)
    > Write option: parallel (default)
    > Size: 6289353 blocks
    >
    > d71: Submirror of d91
    > State: Okay
    > Size: 6289353 blocks
    > Stripe 0:
    > Devic Start Block Dbase State Hot Spare
    > d51 0 No Okay
    >
    > d51: Soft Partition
    > Component: c1t0d0s6
    > State: Okay
    > Size: 6291456 blocks
    > Extent Start Block Block count
    > 0 20971522 6291456
    >
    >
    > d81: Submirror of d91
    > State: Okay
    > Size: 6289353 blocks
    > Stripe 0:
    > Devic Start Block Dbase State Hot Spare
    > d61 0 No Okay
    >
    > d61: Soft Partition
    > Component: c1t1d0s6
    > State: Okay
    > Size: 6291456 blocks
    > Extent Start Block Block count
    > 0 20971522 6291456
    >
    > #############################################################
    > ******************** partition info *******************
    > #############################################################
    >
    > 0. c1t0d0 <SUN36G cyl 24620 alt 2 hd 27 sec 107>
    >
    > Part Tag Flag Cylinders Size Blocks
    > 0 root wm 2904 - 3629 1.00GB (726/0/0) 2097414
    > 1 swap wu 0 - 2903 4.00GB (2904/0/0) 8389656
    > 2 backup wm 0 - 24619 33.92GB (24620/0/0) 71127180
    > 3 usr wm 3630 - 5081 2.00GB (1452/0/0) 4194828
    > 4 var wm 5082 - 6533 2.00GB (1452/0/0) 4194828
    > 5 home wm 6534 - 7985 2.00GB (1452/0/0) 4194828
    > 6 unassigned wm 7986 - 24260 22.42GB (16275/0/0) 47018475
    > 7 unassigned wm 24261 - 24619 506.42MB (359/0/0) 1037151
    >
    >
    > 1. c1t1d0 <SUN36G cyl 24620 alt 2 hd 27 sec 107>
    >
    > Part Tag Flag Cylinders Size Blocks
    > 0 root wm 2904 - 3629 1.00GB (726/0/0) 2097414
    > 1 swap wu 0 - 2903 4.00GB (2904/0/0) 8389656
    > 2 backup wm 0 - 24619 33.92GB (24620/0/0) 71127180
    > 3 usr wm 3630 - 5081 2.00GB (1452/0/0) 4194828
    > 4 var wm 5082 - 6533 2.00GB (1452/0/0) 4194828
    > 5 home wm 6534 - 7985 2.00GB (1452/0/0) 4194828
    > 6 unassigned wm 7986 - 24260 22.42GB (16275/0/0) 47018475
    > 7 unassigned wm 24261 - 24619 506.42MB (359/0/0) 1037151
    ----------------------------------------------------------------------------
    ----------------------------------------------------------------------------
    --------------------------------------------------------
    Thanks to Darren Dunham and Ed Skolnik for pointing me to the exact problem,
    which was that the soft partition was larger than the mirrored meta device.

      The Real Credit goes to Bruce Perttunen ( my colleague here at Blue Cross
    Blue Shield ). Some of Bruce's servers were getting the same ufsdump errors
    as mine. A "Bug Track" is in the works with Sun Microsystems for this issue.
    Bruce sums it up the best and here are his comments and the fix:

    There is a possible file system level problem with soft partitions which are
    mirrored. Depending on how they were created, the filesystem could end up
    being bigger than the metadevice it is sitting on. The only problem this
    has caused to this point is the failure of tape backups of the soft
    partitions, however it is not known if other problems could arise so these
    problems should be addressed quickly.

    Mirrored Soft Partitions - Filesystem Problems

    To determine if there is a problem

    For each filesytem which is built on a mirrored soft partition, run the
    metastat command (where d91 is the mirrored metadevice) to get the size of
    the metadevice.

    # metastat d91
    d91: Mirror
        Submirror 0: d71
          State: Okay
        Submirror 1: d81
          State: Okay
        Pass: 1
        Read option: roundrobin (default)
        Write option: parallel (default)
        Size: 8386767 blocks

    Now get the size of the filesystem by running the following command :

    # fstyp -v /dev/md/rdsk/d91 | grep ncg
    ncg 91 size 4194304 blocks 4128766

    Double the value after the word "size" and compare it to size of the
    mirrored metadevice. If the filesystem is bigger than the metadevice, the
    filesystem must be rebuilt.

    For example :

    4194304 * 2 = 8388608

    Since 8388608 is bigger than 8386767, there is a problem.

    To fix the problem

    In order to fix the problem, the filesystem must rebuilt on the mirrored
    metadevice. To fix the example above, the following steps would be followed
    :

    1. Shutdown all applications or processes using the filesystem
    /dev/md/rdsk/d91
    2. Backup the filesystem /dev/md/rdsk/d91
    3. umount /dev/md/rdsk/d91
    4. newfs /dev/md/dsk/d91
    5. mount /dev/md/rdsk/d91
    6. Restore the data to the filesystem

    If you have problems backing up d91, you may need to break the mirror and
    backup either one of the sub-mirrors or one of the soft partitions.

    What causes this problem

    The problem seems to be caused if a filesystem is created on a soft
    partition and then it is mirrored. For example, the following would create
    the problem...

    1. metainit d20 -p c1t0d0s7 1g
    2. newfs /dev/md/rdsk/d20
    3. metainit d30 1 1 d20
    4. metainit d40 -m d30

    The correct way to do this would be...

    1. metainit d20 -p c1t0d0s7 1g
    2. metainit d30 1 1 d20
    3. metatinit d40 -m d30
    4. newfs /dev/md/rdsk/d40

    Please keep in mind

    If a server is using soft partitions which are not mirrored there is no
    problem. But if at a time later on these partitions need to be mirrored,
    then the filesystem will need to be created on the mirror metadevice after
    the mirror is set up.

    Kirk Gilliam
    _______________________________________________
    sunmanagers mailing list
    sunmanagers@sunmanagers.org
    http://www.sunmanagers.org/mailman/listinfo/sunmanagers

  • Next message: Homan, Charles (NE): "SUMMARY: (NO SUMMARY) Setting serial line receive buffer size?"

    Relevant Pages

    • Re: Solaris 9, SVM soft partition mirroring in a sun cluster set up
      ... Do we mirror the soft partitions the underlying meta devices ... setname = your setname ... or striped metadevice and use that one for creating soft partitions. ...
      (comp.unix.solaris)
    • Re: Solaris 9, SVM soft partition mirroring in a sun cluster set up
      ... Do we mirror the soft partitions the underlying meta devices ... You cannot do the mirror directly. ... setname = your setname ... or striped metadevice and use that one for creating soft partitions. ...
      (comp.unix.solaris)
    • Re: Solaris 9, SVM soft partition mirroring in a sun cluster set up
      ... You can't do this until you've remounted from the mirror ... I have previously mirrored veritas volumes that have single plex to migrate data to new devices.Once the mirror is done, old plexes from the old array are blown away. ... I am under the impression that if I attach the existing sub mirror first to the mirror device, data will be synced from the existing sub mirror to the new ... Long-term I'd prefer to have the soft partitions built ...
      (comp.unix.solaris)
    • Re: Solaris 9, SVM soft partition mirroring in a sun cluster set up
      ... recreate the soft partitions ... appears that I can mirror two soft partitions that are built on two ... built on already mirrored meta devices. ... file systems from the new partitions. ...
      (comp.unix.solaris)
    • SUMMARY: Fixing misconfigured mirror
      ... I had a misconfigured mirror metadevice under Solaris 8/DiskSuite - ... submirror d25, leaving only the single submirror d15 (which was the ... case that it was the physical device that *was* mounted directly that you'd ...
      (SunManagers)