SUMMARY:AdvFS logger panic

From: Pulkkinen Miika <miika.pulkkinen_at_icl.fi>
Date: Fri, 07 May 1999 09:11:00 +0300

Thanks to:

Dr. Tom Blinn
Dale Inman

The problem was advfs-log being full and OS-version/patch-level being too
inadequate to handle the problem. I was able to use salvage to save one
database, which it did execellently(15GB). But the the other database had to
created from backups. I wonder whether or not it would have been possible to
mount the filesystem without the advfs-logger (perhaps mount -d) and the vdump
the whole thing to tape.

These are the answers I got:
You have to boot to single user mode, then make your root file system writable
(mount -u /), then cd into the /etc/fdmns directory and REMOVE the directory
that defines the file domain. Or you MIGHT be able to move the directory into
some other directory hierarchy -- but you must get rid of the domain directory
to assure AdvFS will not attempt to access the domain.

/etc/fstab does NOT reference domains, it only references filesets (although
it references domains by implication).

As you note, the domain is apparently BADLY damaged. There may be no way to
access it without panic-ing the system.

Tom

 Dr. Thomas P. Blinn + UNIX Software Group + Compaq Computer Corporation
  110 Spit Brook Road, MS ZKO3-2/W17 Nashua, New Hampshire 03062-2698
   Technology Partnership Engineering Phone: (603) 884-0646
    Internet: tpb_at_zk3.dec.com Digital's Easynet: alpha::tpb
     ACM Member: tpblinn_at_acm.org PC_at_Home: tom_at_felines.mv.net

  Worry kills more people than work because more people worry than work.

      Keep your stick on the ice. -- Steve Smith ("Red Green")

     My favorite palindrome is: Satan, oscillate my metallic sonatas.
                                         -- Phil Agre, pagre_at_ucsd.edu

     Yesterday it worked / Today it is not working / UNIX is like that
          -- apologies to Margaret Segall

  Opinions expressed herein are my own, and do not necessarily represent
  those of my employer or anyone else, living or dead, real or imagined.




1) DEC support gave us a switch-log procedure that let us increase the value
of the log file. This was to fix ADVFS logger crashes that we had
experienced

2) How often do you defragment your file systems? I have found that a once
a week defrag avoids many problems on the advfs file system.

3) There are parameters that you can set in /etc/sysconfigtab for ADVFS file
systems. Depending on the activity that your system has, these parameters
can make a difference in the health of your filesystems. I would get with
Compaq support and see what they suggest....

Dale Inman
UNIX SYSTEMS
Aladdin Industries LLC
Nashville TN USA

PROCEDURE FROM DEC FOR ENLARGING LOG FILES

       PROBLEM:

        Under some circumstances, AdvFS can panic with a message "log half
        full". At this time, the following situations are known to cause it:
        1) When a very large file truncate is performed (this can occur when

            a file is overwritten by another file or by an explicit truncate

            system call), and the fileset containing the file has a clone
fileset.
  2) When very large, highly fragmented files are migrated
            (this occurs when running the defragment, balance, rmvol, and
            migrate AdvFS utilities).

       RESOLUTION/WORKAROUND:

        In both cases you can work around the problem by reducing file
        fragmentation or by increasing the size of the log.

        You can avoid overflowing the transaction log by increasing its
        size. You can set the log page size when you make the domain using
        the '-l' option ('mkfdmn -l pages', default is 512).

        If you have a spare partition, you can do the following before
        running the commands mentioned above:

        # addvol /dev/rz10b domain
         # showfdmn domain

                         Id Date Created LogPgs Domain Name
         31b8a083.00049136 Fri Jun 7 17:34:59 1996 512 domain

           Vol 512-Blks Free % Used Cmode Rblks Wblks Vol Name
            1 401408 0 100% on 128 128 /dev/rz11b
            2L 262144 192 100% on 128 128 /dev/rz3b
            3 393216 0 100% on 128 128 /dev/rz10b
               ---------- ---------- ------
                  1056768 192 100%

        # switchlog domain 3
        # switchlog -l 1024 domain 2
        # rmvol /dev/rz10b domain

        The above steps add a spare volume to the domain; moves the log to
        that volume; moves it back with an increase in size; removes the
        spare volume. The larger logsize takes up more space on your volume
.

****************************************************************************

These problems by and large have been solved in Digital Unix 4.0E with patch
kit 1.
Recomend upgrade to 4.0E

Thanks
Girish
girish_at_alf.dec.com
-----Original Message-----
From: Pulkkinen Miika [mailto:miika.pulkkinen_at_icl.fi]
Sent: Wednesday, May 05, 1999 4:18 AM
To: tru64-unix-managers_at_ornl.gov
Subject: AdvFS logger panic?


Our Alphaserver 8400 (twin processor) started yesterday telling a
mysterious:
"log_denorec_nunpin: lgr_write failed" of a one adfvfs-domain. Today it has
crashed several times and it keeps crashing during boot (runlevel 3) even
though I have removed the abovementioned domain from /etc/fstab. Compaq told
us
that some internal logs of advfs are come full and our os-version (Dgital
Unix
4.0b patch lev. 8) has some bugs, which cause it to crash. They recommended
that I remove the domain and mayde go to patch lev. 9. I would like to ask,
has
anyone ever encountered similar error and been able to handle it in any way.
Every attempt to mount or use verify to that domain crash the computer.

Best Regards,

Miika Pulkkinen, ICL Data Oy

-------------------------------------------
Miika Pulkkinen, ICL Data Oy
+358-9-5672024, +358-40-7035195
-------------------------------------------

- In a world without walls and fences who needs windows and gates?
Received on Fri May 07 1999 - 06:11:30 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:39 NZDT