Hmm, 14 parts and 14 "Error writing to fd 5" messages. From my memory and a brief look at the 2.6.1 sources (I couldn't find a version in the thread, but this looks like 2.6.1 to me), that wouldn't have come from Amanda itself, but from something Amanda ran - perhaps the decompression binary? You said you disabled custom_compress - is this an uncompressed dump?
The problem is this: all of the parts are concatenated onto the same file descriptor - so if the descriptor (a pipe to amandad in this case) gets closed prematurely, and amidxtaped somehow failed to notice this, then it would make sense to see the remaining parts trigger the same failure. The logs you posted do not have five tapes, but your original problem statement said "when it gets to the 5th tape".. are these two different failures? Dustin -- Open Source Storage Engineer http://www.zmanda.com
