I just started working on an account which is using amanda to do some of its backups of unix servers. I have one backup set where a certain Sun server fails every day with the message 'sendbackup: index tee cannot write [Broken pipe]' and there is a core dump in /tmp/amanda on the client. The rest of the servers in this disklist backup very well. It is trying to use ufsdump for the dumps. The client disks are just slices on a disk (no disksuite or Veritas volume manager). When I look this message up in Google it suggests that the tape is full, but other backups continue writing to the tape.

This Sun server had it's disk reloaded after it experienced a problem. At first, I thought the problem was that ~amanda-user/.amandahosts was owned by root instead of amanda-user. I corrected that, but still the same failure.

I just did a chmod 644 on the raw partitions for the disks (was 640, but amanda-user was in the group that it was owned by) I am trying to dump and will see what that does for tonight's backup.

There are no messages in /var/adm/messages on the client or the amanda server.

Does anyone have a suggestion as to how I can debug this and fix it?

Thanks in advance,

Joe Harpole, Unix System Administrator
IBM Global Services
Phones Voice Fax
(314) 234-8430 (314) 233-8424
Sky-tel Pager: 8883129227 (3129227 AT skytel DOT com)
Cell Phone/Pager: 1-618-670-9032
Web Paging: http://messagings.sprintpcs.com/textmessaging/compose?phoneNumber=6186709032