[mod.computers.vax] More follies w/ volume sets

MHJohnson@HI-MULTICS.ARPA (Mark Johnson) (01/24/87)

  I had an `interesting' experience with volume sets that I ended up
sending an SPR on that I thought others would benefit from.  I started
by performing an image backup of our three disk set to another set of
disks in preparation of splitting our user community to the two volume
sets.  After doing that, the mount command failed with the duplicate
volume name error message.  I mounted the disks privately, used SET
VOLUME/LABEL to change the volume names (seems reasonable...).  Now
mounting the disks I get the message `duplicate volume number already
mounted'.
  At this point I am somewhat confused and call the support center.
They recommend changing the volume set name in VOLSET.SYS.  Arcane
actions with COPY, ANAL/RMS/FDL, CONVERT/FDL, and my favorite editor got
that changed w/ no difference.  I noticed that the INDEXF.SYS file on
each disk (as seen through dump) had the volume set name as part of the
home block (2nd block of INDEXF.SYS).
  To change the file, I tried to use PATCH/ABSOLUTE/NONEW_VERSION.  It
does not work since the file is already open for write (by the system).
I wrote a program to patch it manually (combination of direct access I/O
in Pascal and use of the debugger), dismounted the disks, and found that
it now complained about the bad home block!  I wrote a second program
that used the physical device to recompute the checksum (it is a simple
word by word addition, ignoring overflow) and update it to finally get a
volume set I could use.

  The outcome of all this mess was an SPR for the addition of a command
like SET VOLUME/VOLUME_SET_NAME=xxx to change the volume set w/o this
long and difficult process.  Alternatively, if I had thought to use
BACKUP/NOINIT/IMAGE the volume names would have been ok as well, but
that requires forethought.  If you are using volume sets, please be
aware of this problem when you expand your system and decide to make use
of the extra space by dividing your user community into the two volume
sets.
  --Mark <MHJohnson @ HI-MULTICS.ARPA>