I have been using Amanda (version 2.4.1p1) for about three years as the tool
with which I backup my servers. Currently I backup an NCR 4400 server
(MP-RAS 3.02) to an internal DLT-7000. In the past I had experienced
infrequent errors (vxdump returned 3), but recently a puzzling anomaly has
surfaced.
Here is an extraction from the mail report:
*** THE DUMPS DID NOT FINISH PROPERLY!
*** A TAPE ERROR OCCURRED: [[closing tape: I/O error]].
*** PERFORMED ALL DUMPS TO HOLDING DISK.
THESE DUMPS WERE TO DISK. Flush them onto tape DAILY07 or a new tape.
Tonight's dumps should go onto 1 tape: DAILY08.
FAILURE AND STRANGE DUMP SUMMARY:
<server> /online lev 0 FAILED [out of tape]
taper: FATAL syncpipe_get: w: unexpected EOF
This has happened five times over the past two months. I have researched the
logs for some pattern, expecting that another cron might be interferring,
but that is not the case.
The only common factor I have been able to identify is that the error always
occurs when trying to write a file to the tape which would cross the 2GB
boundary. (My holding disk is 5GB and I'm using DLTIV tapes in a DLT7000
drive. A full backup takes about 18GB of tape space.) It appears as though
Amanda thinks the 35GB tape will hold only 2GB.
The manner in which Amanda sequences filesystems for backup means that it
frequently takes a different number of filesystems (disks ala. amanda) to
reach the 2GB boundary, and they are not always the same filesystems. But
the error, if it happens, always occurs if the next file would cross the
2GB boundary.
Has anyone else experienced this 2GB hiccup?
If so, what is the cause, and how was it resolved?
Any help would be greatly appreciated.
Thanks,
Sid Wood
[EMAIL PROTECTED]
