On Jul 15, 2013, at 4:22 PM, Stefan G. Weichinger wrote: > Am 15.07.2013 23:04, schrieb Stefan G. Weichinger: > >>> How about this (just guessing): Flush- threshold-scheduled 100x (# >>> tapes you normally use) Flush-threshold-dumped. Same number. >>> Taperflush 0 >>> >>> My interpretation would imply that it waits to start taping till >>> there is enough dumped to fill normal tapes. But when all dumps >>> have finished, the taperflush 0 says "send it all to tape; don't >>> save any for the next run". I don't know if Amanda will interpret >>> it that way, but it's what I would try. Deb Baddorf Fermilab >> >> Thanks for your suggestions ... I will test them as soon as I get >> some time to continue working on that server. > > Additional info: > > I don't want to have the behavior "only start writing to a tape if we > have more than X % of tapesize to write". > > I simply want to serialize duping to holding disk and writing to tape. > > The holding disk(s) (it will be a RAID1 array, RAID0 if needed) should > not be stressed by parallel dumping and taping ... I want to make sure > that writing to tape happens with full speed ... no stop and go for that > tape drive. > > For now I don't care for optimal overall runtime, that server might even > write to tape while people are already at work in the morning (the > backup server is a completely separate box). > > So it would be nice to have some simple boolean parameter for this. And > I assume it would be rather easy to implement for a programmer (which I > am not). > > Thanks, Stefan
There's always a crontab entry like this: amdump config --no-taper ; amflush config Deb
