how reliable is publishing smb mount in rsyncd.conf for auto backup?
I've got a number of branch servers (Debian Lenny) ready to be spread out around the country and a large number of branch users who do not back up their files in any kind of organised manner -- which has lead to data loss in the past.
The servers would run samba to provide a quotaed SMB share for each user, which is stored on the branch server's HDD, all in /home/<user> (one dir under /home for each user).
Then, at night, the server would run a cron job that would rsync the contents of the entire /home directory to /mnt/backup/<branchname> (one dir under /mnt/backup for each branch) on a server (as published in rsyncd.conf) (Lenny as well) in HQ.
The HQ server will have mounted a SMB share on a Windows server to /mnt/backup. The Windows server will be running a backup agent that will make sure the data placed there will get to our tape drive.
1) User's copy of Windows mounts branch server's samba share; user saves his documents to here.
2) Rsync ships changes to HQ server running rsync --daemon
3) HQ server's rsyncd.conf looks like:
4) HQ server has already mounted backup something like this: mount -t cifs //10.1.1.1/BackupShare
5) Windows server shares D:\Backup (containing subdirs Ogdenville, etc.) as BackupShare, and backup software is configured to write D:\Backup to tape each night.
Note that the windows share is NTFS, so I don't think that the FAT timestamp issue should affect me?
Would this kind of scheme work and be rock solid reliable in practice?