On Thursday, September 22, 2011 09:44:31 PM Charles Curley did opine: > On Thu, 22 Sep 2011 19:49:45 -0400 > > gene heskett <[email protected]> wrote: > > Thanks for confirming what I said. Recent work seems to have > > destroyed the planners balancing. IMO, when it finds it has lots of > > room, it 'promotes' too aggressively, but it may be repeating level > > 0's on the smaller dle's, leaving some medium sized ones to expire > > and need a level 0 when the big stuff comes around for a level 0, and > > this seems to disagree with a 5 day cycle. > > I'm not sure this makes your argument. What I didn't tell you is I have > DLEs in the hundreds of megabytes and others in the 10GB+ range. With > that, what I often see is the 10GB+ run as level 0 and everything else > run as level 1 or higher. That explains the 9/22 run in the listing in > my last email. That's the only way to get everything else backed up at > all on that run. So I expect a wide variety in the amount of data that > gets backed up on different runs. > Ditto here, I have 36 DLE's, some might not make 30 megs, but 3 of them are unsplitable and exceed 16Gb for the biggest one. 2 weddings I shot on digital hi-8, and that is about half a gig a minute for the raw files. Typical wedding runs 25 minutes or so.
> Your point on only smaller DLEs being promoted is well taken; I've > noticed the same thing. Rarely, if ever, do my medium size DLEs get > promoted. I don't recall ever seeing the monsters ever promoted. I agree, it IMO, should promote, if it is going to promote, a largest fit, and keep promoting until the next largest fit won't fit, leaving the teeny ones to be used to fill in the list of level 3's when the big level 0's come due, and only allow one of those 10Gb & bigger per nightly run. Make a 200k level 4 out of a dozen or more 75 megger's if it has to. There is no obvious rhyme or reason to how its working at the moment. > As long as we're thinking in terms of tapes I'm not sure I'd call the > system broken. It will back everything new up each run. If that leaves > a tape under-used, I have no problem. Occasionally it needs two tapes. > That's fine also. However, if we ever change the basic metaphor to disk > space, that could change. I am trying to keep from overfilling a 1 Tb drive, by adding a day to the cycle when usage passes 900G for 30 vtapes. I just went from 4 to 5 a month ago. > > > Consider enabling tape splitting. That might use your disk space > > > more efficiently than increasing the "size" of the tape. > > > > As in runtapes >1? I considered that, and may yet. > > Also look at: > > part-cache-type disk > part-cache-dir "/crc/amanda10/part-cache" > part-size 900 mb > part-cache-max-size 900 mb > > Which are per tape type, and "allow-split yes", which is per dumptype. Has this spliting been recoverable? IIRC seeing someone having problems with that a month or 3 back, something about incomplete headers on the 2nd and subsequent parts. I haven't tried it personally. I think I'd druther trust runtapes > 1, where it simply restarts that DLE on the next 'tape'. Full headers etc. That would also have the effect of reducing drive use because of the relatively smaller 2nd tape, which won't be filled up with the next run. We'll see what happens tonight I guess. Cheers, gene -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Reality is for people who lack imagination.
