Paul Bijnens wrote:
> Mike Brodbelt wrote:

> This is really an IO error.  Probably due to hardware problems.

Could it not be due to an attempt to write beyond EOT? I'm reasonably
sure the hardware is OK.

> See in the system logs for more information. (/var/{adm,log}/messages
> or syslog).

I do get some errors in the logs, but I'd assumed they were generated
due to an attempt to write beyond EOT. Amanda is quite specific at the
top of the report:-

FAILURE AND STRANGE DUMP SUMMARY:
  castor     sda10 lev 0 FAILED [out of tape]

The log gets this:-

st0: Error with sense data: Info fld=0x0, Deferred st09:00: sense key
Medium Error
Additional sense indicates Write error
sym53c875-0-<5,*>: FAST-10 WIDE SCSI 20.0 MB/s (100.0 ns, offset 15)
st0: Error with sense data: Info fld=0x7fff, Deferred st09:00: sense key
Aborted Command
st0: Error on write filemark.

These errors happen whenever Amanda generates an out of tape error, and
only then. The largest filesystem by far is sda10, and when it gets a
level 0 with all the other fs's getting level 1/2 dumps, all is OK. When
the total data gets to about 30.5Gb (haven't worked it out exactly) it
seems to be

> Try cleaning the drive.
> Push on the scsi connectors.  Unscrew them/ screw back in.
> Move all power cables away from the scsi cables (cross only at 90°).
> Sacrifice a chicken or a goat (depending if you're on scsi2 or UW scsi).

I'll run a clenaing tape and check the cabling for good measure. I think
I'm up to date with my goat sacrifices. Overindulgence in this regard
upsets the staff.

> tapetype from that version would take a few days (and it will not
> warn about compression on or off).  Upgrading would be nice (after you
> solved your hardware problems).

Yes, I thought it would take a while. I may compile a new version to get
a more efficient tapetype. I'm loathe to upgrade it unless I need new
features, as any time I do that with a package I lose the advantages of
Debian's security tracking.

Mike.