Archival tools

[David Nixon]

I am making tape archives of defunct user
accounts to be retained for at least ten years.
Our regular backup tool 'vxdump' has been used
for this purpose, but I want to make the data
more portable and have been testing other
tools: the obvious candidate is 'pax', but this
has problems with long file names: 'cpio' works
better, but is slow. Neither of these handle ACLs. Has anyone got 'star' working under HP-UX
What other tools might be worth testing?

Thanks,
Dave.

[A. Clay Stephenson · Accepted solution]

Whenever I am asked to do long-term archives (and who knows what the eventual target OS/platform will be in 10 years), I fall back on one of two choices that have been in existence for decades on all UNIX boxes: tar and cpio. While neither of these support ACL's, that is a problem that can be outbushwhacked. All you have to do is a find feeding a getacl command that outputs to a textfile. This textfile consisting of filename/ACL tuples then gets backed up along with your regular data. Upon restore, it should be a trivial exercise to read this textfile and restore the ACL's --- assumming that the target OS supports them.

I would lean towards the Gnu version of tar because it supports files up to 8GiB in size with long filenames --- and in the worst-case, you should be able to find the source for it and compile/link it for your future target OS.

Often, you can tremendously speed up a tar or cpio by reblocking the output via dd before sending it to a tape drive.

[A. Clay Stephenson]

By the way, the other "gotcha" is choosing a medium that will still be available in the future.

[TwoProc]

I agree with A. Clay, and have one more gotcha - most media aren't warranted to last 10 years, most tape solutions warranty only up to 5. You can use them longer, but you'll also start experiencing issues with tapes that aren't readable after that long a period as time goes on and the tapes haven't "moved". I've seen tapes last longer than 5, but those are tapes that get at least some infrequent periodic use (imho only). Of course, if you're not using tape, then you don't have the aging issue, but you've got issues on guaranteeing that the media can be read at all, as formats change. Let's say dual layer dvd is used - are you sure you're going to have a dvd-reader hooked to a server that can read it 10 years from now? If so, I've got a circa 1993 HP 9 track scsi drive in the storage room I can send to you! :-)

[David Nixon]

Thanks for that input. I will using 'cpio',
or 'tar'; when large file support is required.
As for ACls, 'gtar' under
HP-UX doesn't save them, or even issue
warnings. So saving in text format - as
Clay sugggests - is the only option.

The tape media I use are DLT: the
"low cost" VS line has a roadmap, so
the data should be OK there for the next ten
years. It is an interesting point that
safe "storage" of tapes may entail periodic usage.