[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Question on backups using rsync

in response to the whole preceding chain.

Arafangion wrote:
On Tuesday 20 December 2005 23:01, Alvin Oga wrote:

automated backup is worthless for that precise reason about corrupted main
systems and there's hundreds of reasons/problems that causes the
main system or backup system to have bad data rendering either or both

backups should be saved not mirrored ... and it's NOT the same thing

Yes, I agree. This is why I take advantage of rsync always unlinking the files before it updates them, which means I can simply rotate the backups, but rather than copying, I just copy the hardlinks.

Essentially it becomes a poor man's revision control system, allowing me to go to any particular day or month.

Thus, I get the convenience of the automated backup, plus the certainty that 'bad things' on the client doesn't result in 'worse things' on the backup server.

This, however, does translate in your partition eventually having to handle _millions_ of files, and it is ideal to back this mess up from time to time, to guard from corruption. (Remember that although a file may exist in 1000 different places, it's still the _same_ file if only hardlinks are used, and is thus succeptable to deliberate or random failures)

I've always worked with the idea of two different kinds of backups.

1) a copy of the critical files (accounting, databases, spreadsheets etc.) that are needed for day to day operations in the event of corruption or accidental deletion and the like. These are just copies, in my case, of just a couple of directories. I don't need long term storage, just a few days/weeks of copies that I can refer to incase I blow something.

2) a system backup with a snapshot of the entire system. With hours and hours of configuration and setup on my boxes, Id like to have occaisional "snapshots" of the whole system. Then if I lose a machine (hardrive crash, theft, flood whatever) or blow the system up somehow, I can recreate the whole thing a-new relatively easily. In this case, the actual critical data from above would theoretically already be stored and retrievable somewhere (and usable on any system) and therefore, these snapshots do not have to be done as frequently. Just whenever a major system change happens, or every couple months to include small incremental system creep.

so for case 1 I'd love something offsite, like a dead webpage or something where I can just automatically load these files and leave'em for now. case 2 needs bigger storage probably and maybe something like partitionimage with the files split onto cds/dvds and stored somewhere safe. The infrequency of this case allows less convenient means of storage.

my .02


Reply to: