one of my unix servers started acting up about a month now, and the behavior just keeps getting stranger.
the server in question is a Red Hat AS 3.0, using LVM, with 7 file systems, all e3fs. the server is running SVN.
1st, this system was backing up fine up to about a a month ago. I'm not sure when it started misbehaving, because the first failure was a silent one; the data from other file systems suddenly became unaccessible.
no errors were reported, and no actual failed backups occurred, the other file systems simply no longer showed up in the selection list or backup sets. once I was made aware of this, I started playing with the server, restarting the agent, running test backups, etc.
my first option was to turn on the include-remote option in the export directive. this seemed to work, but any 0-length files caused the backup to abort. I slowly excluded the portions of the directory tree with touched files, and the selection list seemed to have everything included.
unfortunately, a new problem arose with my first test run; the backup collected over 100 gb from a server with 8 gb total file system usage, and managed to compress it onto one DDS-4 20 gb. the only thing I could think that would cause this is a continual zero set, but I can't fathom what would cause an infinite zero set on this server. only valid data is included in this backup, and "du -h /" shows just under 8 gb used.
my next attempt was to create export directives for each of the files systems. I removed the include-remote directive and added export directives for each of the mounted file systems. however, when I accessed these additional exports from the media server, the selection list showed either an empty file system, or only a very small subset of the directories on the server.
I'm toying with the server now, but it looks like I'm between a roadblock and a minefield. suggestions appreciated.
