>on a particular host (Linux 2.4.6-prexy-xfs, RH 7.1 + LVM + XFS, Amanda
>2.4.2p2) sometimes sendsize fails to estimate sizes with xfsdump.
Give the following patch a try. It won't fix anything, but will sort
out the debug file lines so you can tell which process is doing what.
When maxdumps is greater than 1, sendsize forks for each disk to be
processed. That child is responsible for all the work for that disk.
The parent sendsize never does anything else except wait for all the
children to terminate. And the child is only doing one estimate at
a time. So it's unlikely this is a race condition in sendsize.
>How can I debug this weird thing?
Sendsize may be run by hand. Look for an amandad*debug file with
SERVICE sendsize. Take all the lines in the packet from the OPTIONS
through the last DUMP line and put them in a temp file someplace.
Then run sendsize by hand **as the Amanda user** with that file as
standard input.
You might try setting maxdumps to 1 in the OPTIONS line of the temp file
to see if that makes any difference. You might also try only listing
the file system that shows the error.
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
sendsize.diff
