On 03/05/12 12:40, Jean-Louis Martineau wrote: > What is the value of 'taper-parallel-write'? Set it to 2.
It was not specified, and had defaulted to 1. I had managed to find that in the docs last week and changed it to 2 as you suggest. And yesterday our offsite run worked quite well. I was going to follow-up to my own email, but you beat me too it. :) With taper-parallel-write set to 2, both tape drives were used. Here is the behavior I saw: The backup starts out by mounting tapes in both tape drives, but only writes to a single tape drive until the size of the *completed* dumps exceeds the capacity not yet consumed on that single tape drive. (Add the size of the dumps already taped to the size of the dumps waiting to be taped, once that exceeds the size of a tape, then the second tape drive starts.) The size of in-progress dumps are not included in this calculation. Once two tape drives are being used, it seems that regardless of the size of dumps left to be taped two drives will continue to be used. I ended up with 7 tapes instead of the usual 6 when writing to tapes one at a time. Here is the tape usage report: USAGE BY TAPE: Label Time Size % DLEs Parts offsite-003 2:24 97405M 101.8 51 53 offsite-004 2:24 97476M 101.7 24 27 offsite-005 2:23 97452M 101.6 7 14 offsite-006 2:30 97414M 101.8 55 62 offsite-007 2:27 97240M 102.3 165 173 offsite-008 0:40 27424M 28.6 1 4 offsite-009 0:45 31128M 32.5 0 4 When tape offsite-009 was started, all the dumps were complete, and what was written to offsite-009 would have fit on offsite-008. I can see it being nice to use fewer tapes, but I'll use an additional tape if I can complete the backup faster. I actually would have preferred that Amanda would have started right off using both tape drives. I suppose this could be a knob: taper-parallel-write-optimize and settings for "speed" and "space". The first starts writing to taper-parallel-write tapes right away; and the latter doesn't use any more tapes than needed. Right now the knob is set to "neither" it seems. :) Amanda 3.2.2. Thanks for the tip Jean-Louis.
