Mike Fedyk wrote:

Jon LaBadie wrote:

On Tue, Aug 31, 2004 at 04:57:42PM -0400, Jason Castonguay wrote:

Right now I am using amanda for a few different linux servers. My disklists use the directories where different partitions are mounted. This has been working pretty well, but I read somewhere that one can get even better results by breaking things up to as many small directories as one can.

Not exactly "as small as possible", but small compared to the capacity
of the tape.  When amanda writes an image to tape, and bumps into end
of tape, it will restart the complete image again on the next tape;
the last, incomplete, backup image on the previous tape is lost
capacity and lost time too.

When using the directive "taperalgo largestfit", amanda starts to write
the largest image that will fit on the tape (supposing you got
reasonable numbers in the tapedefinition).  If, between the dumps on

EOT, but now at least the lost time and capacity is minimal.  (There
exists a patch that chooses the smallest because that one has best
chance to fit completely on the tape.)

Obviously, the algorithm works better if you have many files to choose
from, instead of only 1 large file.


>>>               This makes sense to me, as it will allow

amanda to schedule different level backups the best she can even if it makes things perhaps more difficult to keep track of.

So I collected on each client the names of the harddrives for it.

find / -type d -fprintf ~/amandalist "`hostname -f`\t\t%p\tcomp-user-tar\n"

I'm afraid that you fall into the other extreme with this fragmented
setup.  Maybe you even hit the maximum size of a dgram (64Kbyte if
the underlying OS can handle it) that amanda uses in the estimate
phase.  Not to count the performane loss to setup/start/stop all
the subprocesses etc.

And if you ever have to restore a complete disk (that's one of the
main reasons why you make a backup, isn't it?), you'll have to restore
umpteen tiny little directories.  Restoring backups is usually done
in a highly stressful situation, and I doubt you'll see the humor in
typing umpteen restore commands while people keep asking how long it
will take.

and I have an idea to exclude all subdirectories in the dumptype, but then I run into the problem of users creating and removing their own subdirectories and then dealing with those. Do people who advocate this sort of setup generate their disklist each time they run amanda? or do they not have to deal with this problem? What is the best use of gtar?

Look for a few large subdirectories, and make a separate
DLE for those.  Then group the other ones in meaning parts using
globbing (only the one level deep!!!).  Make sure the globs cover
all names, even the one that don't exist yet.

See the last example in examples/disklist in the source tarball.



--
Paul Bijnens, Xplanation                            Tel  +32 16 397.511
Technologielaan 21 bus 2, B-3001 Leuven, BELGIUM    Fax  +32 16 397.512
http://www.xplanation.com/          email:  Paul.Bijnens AT xplanation DOT com
***********************************************************************
* I think I've got the hang of it now:  exit, ^D, ^C, ^\, ^Z, ^Q, F6, *
* quit,  ZZ, :q, :q!,  M-Z, ^X^C,  logoff, logout, close, bye,  /bye, *
* stop, end, F3, ~., ^]c, +++ ATH, disconnect, halt,  abort,  hangup, *
* PF4, F20, ^X^X, :D::D, KJOB, F14-f-e, F8-e,  kill -1 $$,  shutdown, *
* kill -9 1,  Alt-F4,  Ctrl-Alt-Del,  AltGr-NumLock,  Stop-A,  ...    *
* ...  "Are you sure?"  ...   YES   ...   Phew ...   I'm out          *
***********************************************************************