HallMus

Im sysadmin on a regional museum, länsmuseet Halmstad http://www.hallmus.se

We have switched to a pure HD-backup solution.

First we tried to run everyting with dirvish - it worked, but it was a rather wastefull on disk.

Then about a year ago we started combining Dirvish and backuppc - that has turned out to be a much better solution!

Dirvish excells at being very low on CPU, the drawback I had with Dirvish was that if I had multiple systems with same file(s) it was saved once for each system, making the disc-consumption unrealistic.

Enters backuppc. This app is a very good complement to Dirvish, its good at exactly what isnt optimal in Dirvish and vice versa.

backuppc is also a rsync-based system, but there the actual stored file is named after its md5sum and compressed, each filetree has links to the data-tree where all the md5sum files are stored. This gives that a file, regardless of filename is only stored once.

This saves so much backupspace, so we actually backup full disks, not bothering with include/exclude filters (except for /var/log) - We backup some 20 windows boxes and 10 linux servers - storage required is 160GB for a rolling daily three month backup

The drawback with backuppc is its enourmus CPU usage, on files comming from just one source it is also totaly useless overhead - no gain since no other source will have the same files.

This is where Dirvish enters. Since we do a lot of digitizing of images, and the originals are stored on a single server - that dataset has no advantage of a backuppc solution, so for those files we use Dirvish, storage required 100G also a roling threemonth backup, this we run 2 times a day.

For redundancy we have two systems using the Dirvish+backuppc setup, located in different buildings.

Since the images are kinda important, we also run the Dirvish part of the backup on two aditional servers of-site.

Then each quarter we just change the destination for the backups to external drives with no expire to get permanent backups. The external drives are normaly stored of-site, only connected during the quarterly backup, then imeditally removed.

We have two sets of external drives A and B, we use them alternating quarters. When one external drive is full, its retired and stored.

To sum it up - Dirvish didnt solve all our backup needs, but it definetly solved a large part of it, and since its so low on bandwith/CPU it can run at daytime without disturbing our LAN.

In my opinion Dirvish + backuppc is an excelent combination!

HallMus (last edited 2011-01-24 05:15:27 by KeithLofstrom)