Our backup system seems to be working...
It works like this: any time a backup is made, it's scanned by a script and all the MD5s are put in a database. This also checks that the backup is readable at all. Semi-regularly, perhaps once a week, another script goes through the entire filesystem, computes MD5s of all files, and puts those that aren't in the database into a list, in convenient 4.4GB chunks (or 680MB, if preferred) and convenient format (used to be mkiso file list, now it's a script that copies the files to a temporary location). We then burn these and go back to the beginning of this paragraph.
Right now we have 14 of them scheduled.
One neat feature is that if we make a mistake - skip a chunk, say - it doesn't matter. On the next run through, it will schedule those files again, until they get into the database.
This will be put up on SourceForge when I clean it up a bit; drop me an e-mail if you want it earlier...
⇦ All the taxes paid in a lifetime | ⇨ Idea: mat2mat protocol |
