On Sat, 27 Apr 2002 at 9:15pm, James Long wrote

> I'm brand new, so let me first say thank you, this looks like a pretty
> full-blown backup implementation.

Welcome.

> Bunch of new user questions: can I put the holding disk on one
> of the file systems I'm backing up (backing up the tape server host
> itself)?

Yes.  You can either specify the holding directory in an exclude list or 
specify 'holdingdisk no' in the dumptype for that fs, which will tell 
amanda to dump that filesystem straight to tape.

> Does amanda backup the entire filesystem, both the used space and
> the empty space?  Since it uses the fs /dev/ entries, do I need to
> be concerned about symlinks that cross slices?

This is something that confuses a lot of new users.  Amanda schedules 
backups, it doesn't actually get the bits off the disk.  It uses either 
dump or GNUtar to do that.  How they handle symlinks, etc, are how they 
will be handled.  With either tool, btw, only data gets backed up.

> I'm using a Seagate Scorpion DDS-2 (4G/8G) drive.  Should I skip amanda's
> compression and use the hardware compression?  I've been using tar, and

This depends.  Software compression puts more load on the machines getting 
backed up, but let's you more accurately see how much data is going on 
tape.  Hardware compression takes load off the systems, but makes you 
guess as to how much compression is going on (and thus how "big" your 
tapes are).

> I made uncompressed tar files of about 1.8G and 1.2G and used hardware
> compression to get them to both fit on a 2G/4G tape.   But I've added
> more content, and now the 1.8G file is over 2G, and tar seems to use a
> signed 32-bit int, so I'm hosed beyond 2G (.tar cannot contain any single
> file over 2G).  So since the tape is a .tar file itself, it was crapping
> out on the 2G file.

Amanda writes the tar files to holding disk in chunks of a size specified 
in your amanda.conf (chunksize).  This is to get around the 2GB limit.  
When going directly to tape, that limit doesn't exist.  So tar is fine.

-- 
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University

Reply via email to