> On Mon, Apr 22, 2002 at 11:39:55AM +0200, Toralf Lund wrote: > > On 19/04 2002 16:18 Dave Sherohman wrote: > > > You can do this manually by not changing tapes (or leaving the tape > > > drive empty) tonight, which will cause all of tonight's dumps to stay > > > on the holding disk (provided it's big enough), then put the new tape > > > in and run amflush tomorrow morning to collect all of the untaped > dumps. > > > Yes, of course. Does this also mean that it's a good idea to use the > > amount equivalent to one tape of the holding disk space? > > IMO, holding disk space is a matter of 'the more, the merrier'. > You definitely want enough to handle at least one runs' worth of level > 1 dumps, in case someone forgets to switch tapes or the taping never > starts for some other reason. I like to go with at least twice that > (disk is cheap, these days...) so that I can consolidate partial dumps > if something goes wrong mid-run. I was thinking that allowing more than one tape's worth of data to be stored on disk would be bad as there might still be something left after flushing an overflowed dump + run without tape, which would leave me in exactly the same state as before I started.
> > > Nope. Last night's untaped dumps will sit on the holding disk > forever > > > unless you either delete them or run amflush > > Too bad. > > Yeah, it would be nice if you could set amanda up to automatically do > an amflush onto the same tape after completing a normal run if the tape > isn't full yet. Exactly. Or to do amflush first when starting the next run, then do new dumps to fill up the tape. > > > I'm assuming there is no way to continue writing to the tape were > > something was amflushed, either, but please let me know if I'm wrong. > > You are correct. Amanda will not append to tapes under any > circumstances. > It always starts writing from the beginning of the tape. Wouldn't append support be easy to implement? Seems to me that most of the code must be there already (since Amanda writes several dumps to one tape in its normal mode of operation.) - Toralf
