[comp.sys.att] Backups to 8mm under SysV

cgh018@tijc02.UUCP (Calvin Hayden ) (01/03/90)

Well, here goes...
I have posted on this before, and am now looking for further
input.  We have a Vax 8600 running a port of AT&T Sys V R2 V2.
We recently got a couple of 8mm tape units from MicroTechnology.
The ultimate goal is to implement unattended backups using these, 
to allow us to pull our third shift operator to first.  The tape 
units emulate a TU81, and are using the tmscp driver.  Currently,
we do daily, weekly, and monthly backups.  A daily is done via a
"cd some_fs;find . -mtime -1 -print|cpio -.....", a weekly is done
the same - with 8 days instead of 1, and a monthly is done by using
volcopy to copy the entire file system to tape.  Here is the information
I have gained through 3-4 weeks testing:

* I can do a daily or weekly with no problem.  cpio seems to work
  with or without the B option, and I get no error messages either
  reading or writing the tape.

* When using volcopy on any unmounted file system, I get a console
  error message something like "tms1a:  hard error :  record data truncated",
  and sometimes "tms1a: hard error: serious exception".  The error occurs
  at the beginning of the volcopy, but not during the write or read phase.

* When using volcopy on a small file system (say 200,000 blocks) :-), the
  above message occurs.  I continue, and then restore from tape to disk.
  A fsck of the original and the restored copy shows the same, and a dircmp
  of the two directories shows the same.  Mounting and using the copy seems
  to work too.

* When using volcopy on a big file system (say 1,000,000 blocks), I get the
  above error messages, and continue.  However, when I try to restore the
  file system, I get read errors, and the above type errors on the console.
  The restored copy is no good.

Given the above, I feel secure with using cpio, but not volcopy.  I am looking
at using cpio to do monthly backups, and want input on this.  We have file 
systems that range from 100,000 blocks to 1,190,000 blocks, and all get a
good bit of traffic.  My main reservation about using cpio for the monthly is
the case of restorations.  I have lost a drive/file system before, and it was
simple to restore from the volcopy monthly.  Using a cpio monthly, this would
have been a pain (ie: our cpio wont allow recursive restores such as /usr1 and
all subdirectores and files - and I'm not fond of having to do a restore sub-
directory by subdirectory).  I have played with afio, but dont know if I can
do this with it either - I'll check, since I know afio has better recorvery
than std. cpio - maybe it's better at this too.  I also have reservations  
about "stacking" one file system after another on 8mm since we only have 2
of the units (ie:  Tape1 - usr1 usr2 usr3, Tape2 - abc def ghij kl,...etc)
by using no-rewind when using cpio on the first n-1 file systems to go on tape.

Anyone out there doing large backups to 8mm in any of these ways?  Comments
on the +/- of using cpio or volcopy for monthlys?  If I should have to go
with cpio - any suggestions on the best way to do so? 
Anyone have suggestions as to the above error messages?  Comments on having to 
restore from a cpio archive of large size?  For those who have responded
to earlier questions about backups on 8mm units - THANKS.  To those of
you who respond - THANKS IN ADVANCE.

Additional Information:  We have 10 file systems that get backed up during
 monthlys.  Each file system is volcopied to its own set of 9track tapes in
 our current configuration.  Yes, I know that the file systems are rather
 large, but its an evil I have to live with for now.


Calvin Hayden
Texas Instruments, Johnson City, Tn.    
Voice (615)461-2254
UUCP:  ...mcnc!rti!tijc02!{root,cgh018}