Jon — you are brilliant! The backup SERVER’s dump files, which I did ad hoc after all other jobs were done and just left there sitting on the holding disk, were indeed the first files onto the tape. What a clever idea! I did a force on the ad hoc dump, so the the dump sitting on the holding disk was a FULL dump. Very useful concept. :D
Deb Baddorf Fermilab On Dec 17, 2014, at 12:57 PM, Debra S Baddorf <badd...@fnal.gov> wrote: > > On Dec 17, 2014, at 12:18 PM, Jon LaBadie <j...@jgcomp.com> wrote: > >> On Wed, Dec 17, 2014 at 05:48:07PM +0000, Debra S Baddorf wrote: >>> Expounding on Jon’s ideas: how about, after all the backups are done and >>> finished for the day. >>> Then, I run another job >>> amdump <config> —no-taper <dumphost> <all-its-DLEs> >>> >>> This would leave them sitting on the holding disk, waiting for tonight’s >>> backup job. In fact, I’ll do that right now, manually. >>> Tonight, I’ll see if they happen to flush to tape first. They might >>> not, since I’ve got “pick biggest file” as my taperalgo . >>> But it’s an interesting experiment. >>> >>> Deb Baddorf >>> Fermilab >>> >> >> Would you need to do a level 0 to make the amanda data worthwhile? > > Good point. Maybe yes. The point is that I would want the amanda log files > the most, so I know where data is on my tapes. > I could recreate the boot disks from source but I’d prefer not to do, in a > hurry — so I’d best have boot disks there, too, > so yes, let’s force those to level 0 too. >> >> IIRC, the taperalgo is in effect when multiple DLE's are available >> for flushing. So if it(they) is(are) the only one(s) available ... >> >> You may want to check your settings for autoflush and flush-threshold-*. >> The first says whether stuff on the holding disk should be flushed >> with the amdump run and the latter says how much data must be available >> before starting taping of a new tape. > > I’m pretty sure my flush-threshold-* are already set to “always flush every > last bit”; > don’t save stuff on the holding disk, don't be frugal with tapes. And > autoflush is set > to “all”. Good point. > > Ok — let’s run another —no-taper dump of the one DLE that wasn’t already > done at level 0. > > Deb Baddorf > Fermilab > >> >>> On Dec 17, 2014, at 3:20 AM, Chris Ritson <chris.rit...@newcastle.ac.uk> >>> wrote: >>> >>>> >>>> >>>> At one point I thought it might help disaster recovery if Amanda's own >>>> status information was last on the tape so I used a starttime in one >>>> specific dump-type, picking a time which, at least 90% of the time was >>>> after the overnight run normally finished. I guess if something had to be >>>> first you could use a starttime in everything else (presuming that the >>>> Amanda run starts at a fixed time of course). >>>> >>>>> -----Original Message----- >>>>> From: owner-amanda-us...@amanda.org [mailto:owner-amanda- >>>>> us...@amanda.org] On Behalf Of Debra S Baddorf >>>>> Sent: 16 December 2014 18:09 >>>>> To: amanda-users Users >>>>> Cc: Debra S Baddorf >>>>> Subject: Re: How to priorize DLEs? >>>>> >>>>> I've often wondered (and then dropped the idea) - is it possible force a >>>>> particular DLE X to be first on the tape? >>>>> For instance, it seems useful to have the backup machine's own backup to >>>>> always be first, >>>>> in case of catastrophe. My impression was that it was NOT possible. >>> >>>>> End of included message <<< >> >> -- >> Jon H. LaBadie j...@jgcomp.com >> 11226 South Shore Rd. (703) 787-0688 (H) >> Reston, VA 20190 (703) 935-6720 (C) >
