[re-listed from amanda-hackers]
>Chris, > >This might be more appropriate for amanda-users, but I'll give it a go: > >I don't see how a code change will help your hardware, and there are enough >people on the lists on low- or no-budget networks that I won't suggest a new >tape drive (yet). (Aside: clean the drive if you haven't lately, this sounds >like dirty heads.) > >After each backup, I would suggest using amverify to determine if the tape >is any good, then removing and re-labeling the tape if amverify fails. You >said this 'doesn't appeal', but I think its cleanest when you've destroyed a >good(?) backup on that media with a bad one. You need to get rid of the >indexes for the tape regardless of what you do with the physical media. > >If you write a script to amverify/amrmtape/amlabel, you could make it log >the re-labels to look for a pattern, such as a specific brand of tape >causing the problem. Does anyone have such a script to share? that is a: "script to amverify/amrmtape/amlabel" > >I would also be VERY suspicious of the 'good' backups on a drive that is >sometimes 'bad'. If the data's that important, then a reliable backup drive >should be equally important. > >Best of luck, >Paul > >> -----Original Message----- >> From: Chris Jones [mailto:[EMAIL PROTECTED]] >> Sent: Wednesday, June 05, 2002 2:23 PM >> To: [EMAIL PROTECTED] >> Subject: ...but I didn't use that tape! >> >> >> I have a lousy tape drive. When I try to dump (or flush), >> sometimes the >> backups will fail on the first file (after the label) on the >> tape. In >> other words, there's no valid backup data on the tape. But Amanda >> considers the tape to be used, and I can't put more data on >> it unless I >> either wait for the tape cycle to come around again, re-label >> the tape >> in question, or try and hack on the datafiles. For various reasons, >> none of these options really appeals to me. >> >> Would it be desirable to special-case this situation? If a >> tape write >> fails before the first file has been completely written, >> don't consider >> the tape to have been used. >> >> Chris >> >> -- >> ------------------------------------------------- [EMAIL PROTECTED] >> Chris Jones Mad >> scientist at large >> www.netbsd.org www.postgresql.org www.schemers.org www.python.org >>
