>It did go away by itself when the same thing happened 6-8 months ago.

Well, then, make it do it again!  :-)

>Okay, script gzip.test put in place and amanda run with a disklist
>just for the one troublesome client.  5 partitions.  3 work, 2 fail.
>The gzip.log files don't give me any great hints:

There's a bit more information there.  The two that failed returned a
status of 152 -> 0x98 -> 0x80 + 0x18 -> 0x80 + 24.  Freely translated,
that's still our SIGTSTP signal (24), but it also says gzip dropped core
(0x80).

Take a look in /tmp/amanda and see if there are any core files from
gzip.  Try:

  $ gdb /bin/gzip core
  gdb> bt

I also looked through all the Amanda source and it never sends a TSTP
signal, just TERM and KILL (that was not a surprise).

You might try changing the script to call gzip like this:

  /usr/bin/strace -o /tmp/gzip.strace.$$ /bin/gzip "$@"

If it generates too much output, you might try:

  /usr/bin/strace -e 'trace=!read,write' -o /tmp/gzip.strace.$$ /bin/gzip "$@"

This is fast becoming a gzip or Linux debugging session rather than an
Amanda one.

>  Chris_Marble

John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]

Reply via email to