I have a fileserver with a few hundred gigabytes of data which I rsync to a remote server nightly. With each rsync the old data is hard-link copied to a new directory, so I have daily backups going back several weeks. I then also have an external hard drive at the remote location which does a backup once a month and is disconnected most of the time.

So I think I have my backups covered. The thing that bothers me now is silent corruption on the file server. If a file is corrupted and not caught, it will be synced to the backup server. And given enough time, all the good copies of the file will fall off the end of the backup schedule.

Is there a script that will check for individual file integrity on a whole partition, or a directory tree? Maybe it would create hashes nightly and keep a log of changed files. Is this a reasonable thing to do?

Since I'm primarily concerned with corruption due to hardware and not intentional malicious corruption, would CRC32 be an appropriate checksum? I'm thinking there would be speed issues with MD5 or SHA-1.