On Tuesday 30 October 2018 15:29:37 Nathan Stratton Treadway wrote: > On Tue, Oct 30, 2018 at 14:20:55 -0400, Chris Nighswonger wrote: > > Why in the world does Amanda plan level 0 backups for all entries in > > a DLE for the same run???? This causes all sorts of problems. > > > > Is there any solution for this? I've read some of the creative > > suggestions, but it seems a bunch of trouble. > > The operation of Amanda's planner depends on many inputs, both "fixed" > (e.g. configuration options) and constantly-varying (e.g. estimate > sizes and dump history), and I suspect there are only a few people in > the world who really understand it fully -- and I don't know how many > of them still read this mailing list :(. But even one of those people > would probably need to look at a lot of information in order to know > what exactly was going on. > > > The good news is that I have noticed that the planner records a bunch > of interesting information in the amdump.DATETIMESTAMP log file, so at > least that seems like the place to start investigated. Look in > particular for the following sections: DONE QUEUE, ANALYZING > ESTIMATES, INITIAL SCHEDULE, DELAYING DUMPS IF NEEDED, PROMOTING DUMPS > IF NEEDED, and finally GENERATING SCHEDULE. > > In your case, it seems likely that the PROMOTING DUMPS section should > have a bunch of activity listed; if so, that might explain what it's > "thinking". > > If that doesn't give a clear answer, does the INITIAL SCHEDULE section > show all the dumps are already scheduled for level 0? If not, pick a > DLE that is not shown at level 0 there and follow it down the log to > see if you can figure out what stage bumps it back to level 0... > > > On a different track of investigation, the output of "amadmin CONFIG > balance" might show something useful (though off hand it seems > unlikely to explain why _all_ DLEs would be switched to level 0). > > > Let us know what you find out :) > Nathan > I just changed the length of the dumpcycle and runs percycle up to 10, about last friday while I was makeing the bump* stuff more attractive, but the above command returns that the are 5 filesystens out of date: su amanda -c "/usr/local/sbin/amadmin Daily balance"
due-date #fs orig MB out MB balance ---------------------------------------------- 10/30 Tue 5 0 0 --- 10/31 Wed 1 17355 8958 -45.3% 11/01 Thu 2 10896 10887 -33.5% 11/02 Fri 4 35944 9298 -43.2% 11/03 Sat 4 14122 10835 -33.8% 11/04 Sun 3 57736 57736 +252.7% 11/05 Mon 2 39947 30635 +87.1% 11/06 Tue 8 4235 4215 -74.3% 11/07 Wed 4 19503 14732 -10.0% 11/08 Thu 32 31783 16408 +0.2% ---------------------------------------------- TOTAL 65 231521 163704 16370 (estimated 10 runs per dumpcycle) (5 filesystems overdue. The most being overdue 20 days.) That last line is disturbing. Ideas anyone? I'll certainly keep an eye on it. Cheers & thanks, Gene Heskett -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene>