rdump to sun with Exabyte drive

[lowry at arnor.uucp]

In article <AUVNELE.90Oct24161059 at auvc7.tamu.edu>, auvnele at auvc7.tamu.edu (Eric L. Nelson) writes:
|> I'm trying to do an rdump from an RS/6000 to a sun 4/330 which has an
|> exabyte 8mm drive. No matter what I put on the -l number2 (the -l
|> option specifies the max number of blocks to transfer to the remote
|> media) it always dumps out some of amount and then asks for a second
|> tape. I only have ~400 meg to write so I know I'm not putting more
|> than the tape can hold.
We've had success with:

	/etc/rdump -<level> -l 5000000 -d 300000 -f<host>:<tape> <filesystem>

I don't think there's anything special about our numbers except that
they're both very big.  We ran into two other problems with rdump, as
well:

  1. A storage leak was causing our dumps to fail on very large
filesystems.  Someone here managed to work out a fix to rdump.  I
don't know what was involved, but I'll ask the guy who did it to post
additional information if possible.

  2. /etc/rdump sometimes continues to produce output after it has
terminated.  I imagine this must be one of the many child processes
spawned by rdump.  I don't know why rdump doesn't wait for its
children.  This was causing us a lot of trouble with our backup
scripts that invoke rdump via rsh and then (since rsh does not
propagate return codes), echo $status after rdump finishes.  We were
unable to pick out the return code in some cases because it came out
in the middle of rdump output.  We ended up putting in a 10-second
sleep between the rdump and the echo.

--Andy Lowry, lowry at ibm.com, (914) 784-7925
IBM Research, 30 Saw Mill River Road, P.O. Box 704, Yorktown Heights, NY  10598