FAIL, usually means the back hit a unrecoverable error of some kind and was forced to stop. data backed up before the error would be good, and even after the error if only one server was failed (as opposed to the whole job failing).
if this is a recurring problem specific to one server, consider moving that server to the end of the resource order in the job settings. it would allow you to get more of the data before the failure, or, in the event of the error staying in the same relative place after moving the resource order, eliminate that server as the source of the error.
as mentioned, check the log (or, perhaps better, post the whole error here so we can help).