Something I mention in my book, but which is worth elaborating further upon, is the need to keep backups of your backup server for as long as your longest backups – if not longer. One of the primary reasons for this of course is the indices; recovering older indices is traditionally easier than the laborious alternative of scanning in potentially a multitude of media.
There is however another, equally important reason why your backup server should have at least equally the longest browse/retention time in your site – the logs.
Being able to recover your backup logs (i.e., nsr/logs/daemon*, nsr/logs/messages, etc.) is like having your own personal time machine for the backup system. This becomes important when you hit recovery situations that you just can’t explain. That is, an error you are getting now, when you try to do a recovery of files backed up 2 years ago, may not make any sense at all. However, if you’re able to recover the backup server logs from that period in time, they may very well fill in the missing information for you. The most common thing I find this helps with is identifying whether what you’re trying to recover was ever actually backed up in the first place. I.e., the scenario runs something along the lines of:
- User asks for file from arbitrary date – e.g., 29 May 2006.
- Can’t browse to 29 May 2006, but can browse to 28 May 2006 and 30 May 2006.
- Recover backup server logs from 30 May 2006 to see that the client could not be contacted for backup on that day.
Now, some would argue that not being able to recover is the real problem – this isn’t always the case. Sometimes, due to circumstances beyond your control, you literally can’t recover – such as say a situation like the above where there was a failure to backup in the first place. In situations such as this, being unable to explain why the recovery can’t be facilitated is equally as bad as not being able to recover.