Deb, thanks very much for this detailed answer. So if I understand correctly, if I follow your advice and have a number of entries in my config's DLE, and call 'amdump myconfig', each DLE entry is handled as a separate dump as amdump runs? That is, my main concern is to do dumps in chunks so that with large full dumps, if there's an error, there's no single massive dump that gets compromised and must be repeated. Also, worst-case scenario restoration from smaller dumps will be easier. Does that sound right?
When doing incremental dumps with this setup, does amdump combine the changed files from various entries in the DLE until it reaches the minimum dumpsize I set to fill a tape? Or is each DLE still handled separately even if none of them is large enough to fill a tape (or match whatever dump size limit is set). Also, I like the advice on doing archive dumps, seems straightforward. I'll be needing to do those too. -M On Thu, Jan 16, 2014 at 11:41 PM, Debra S Baddorf <badd...@fnal.gov> wrote: > Since others have diverted into providing answers for your tape type, > I'll resurrect this question and give some sort of an answer. > > Yes, you probably need just one configuration and one group of tapes > labelled per 'labelstr'. You create a list of DLEs to make smaller than > tarp sized chunks. > node /diskname dumpname [spindle Local ] (optional) > node /disk2/aDir tar-type-dumpname [...] > node /disk2/another tar-type-dumpname [...] > node2 /diskname dumpname [spindle Local ] (optional) > node2 /disk2/aDir tar-type-dumpname [...] > node2 /disk2/another tar-type-dumpname [...] > Etc > > Then, optimally, you create a cron job for each night (or maybe only > week nights), something like this: > 21 00 * * * amdump myconfig > And let Amanda shuffle your DLEs into a similarly sized amount each night. > > Stop reading here, for the simplest setup. > > ------------------- > > If perchance the manager of nodeN insists on checking his dumpdates log > on the same day each week, you CAN force nodeN with another cron entry & > script: > > "If Wednesday then amadmin myconfig force nodeN * " > Or perhaps only amadmin myconfig force nodeN /usr > > Things like that. It forces Amanda to do a level 0 THAT day, on > whatever DLEs you do it for. It messes up Amanda's balancing, but if you > have a lot of DLEs it'll even out somewhat. > > You also CAN do a force on all DLEs ( * * ) on the day that you want a > whole set of level 0's. But that really messes up the balancing & isn't > the best idea. > > ----------------- > > I personally DO run a second configuration (archive) and a second set > of tapes (same tape drive) so that I can get an archive tape. I set it "no > record" so it doesn't change the dumpdates, and so it doesn't stop other > level 0's from being done in the normal configuration . I do a "force" on > all my DLEs and run it once a month. These tapes go offsite, so I want my > daily configuration to have its own set of level 0's. Level 0's at least > once a week (my dumpcycle) but then I keep 5-6 weeks of tapes before I > recycle and start overwriting those level 0's. > > > > Deb > > On Jan 16, 2014, at 5:01 PM, "Michael Stauffer" <mgsta...@gmail.com> > wrote: > > Thanks for the reply. I don't have multiple configurations, I'm just > trying to figure out how to set things up. > So it sounds like I can have a single configuration that uses the same > group of tapes as defined by 'labelstr', and I use the DLE lists to break > things up into manageable chunks. Then with amdump, I can specify the > individual DLE's to be dumped? > > -M > > > On Thu, Jan 9, 2014 at 1:58 PM, Jean-Louis Martineau <martin...@zmanda.com > > wrote: > >> On 01/09/2014 01:47 PM, Michael Stauffer wrote: >> >>> >>> Hi, >>> >>> I'm setting up amanda 3.3.4. >>> >>> Regarding 'labelstr' in amanda.conf: >>> >>> The documentation says: "If multiple configurations are run from the >>> same tape server host, it is helpful to set their labels to different >>> strings (for example, "DAILY[0-9][0-9]*" vs. "ARCHIVE[0-9][0-9]*") to avoid >>> overwriting each other's tapes." >>> >>> Does this mean that if I have multiple configurations in order to break >>> up a large 30TB data set into managable chunks, each configuration will >>> have to have a particular set of tapes assigned to it? That seems very >>> awkward if so. >>> >>> >> yes >> >> Why do you have multiple configurations? >> You should do it with one configuration and multiple dle. >> >> Jean-Louis >> > >
