Sun 04 Mar 2007 09:03:02 PM UTC, comment #2:
I see this as a bug for two reasons:
First, proper stdout/stderr separation, redirecting stdout to /dev/null, and using cron to run your backup will get you mailed error messages that only appear in the event of failures. (That's why rsync and traditional UNIX tools like cp work the way they do) I suspect this is where our difference of opinion primarily originates.
Second, I (and many other admins, I'm sure) would much rather see the backup succeed partially than fail in the following situations:
1. The backup depends on proper file permissions and there's a malicious user with enough access to change those permissions. (In my case, I run my backups as a non-root user for the same reason one would chroot an http daemon. Things like /etc/shadow get backed up separately.)
That works in two situations: First, if the user in question has ownership of (or group access to) everything non-public that's being backed up. Second, if the system is using something other than the standard UNIX file permissions. (eg. POSIX ACLs, SELinux security rules, etc.)
2. Something goes wrong (it doesn't matter what) and the admin is away. In this situation, total failure to make backups could lose the admin his/her job.
3. Permissions get mangled by impending drive failure. (I don't know if this is possible, but I doubt you want to risk the wrath of angry admins)
Also, how do I re-open a bug in Savannah? I only have experience doing so with Bugzilla.
|