Re: bump* dumptype parameters tuning

2018-04-09 Thread Jon LaBadie
On Mon, Apr 09, 2018 at 01:20:05PM -0400, Jean-Francois Malouin wrote:
> Hi,
> 
> It's with some shame that I must admit that through the long years of using
> amanda I've never really played with the bump* parameters for DLEs!
> 
> I'm now in a situation that might benefit from some tuning using those
> to aggressively encourage bumping the gnutar incremental levels.  The
> drawback of potentially needing more tapes for recovery is not really a
> constraint for me.
> 
> I have a DLE (~5TB) for which the estimates are done on the server --
> the client, a NAS/filer, doesn't support anything else, so the DLE usage
> info is really just from the historical usage not the current one. Any
> ideas/hints on how should I proceed from what is currently (default
> values I guess) configured?
> 
> bumppercent 20
> bumpsize20480K
> bumpdays1
> bumpmult4.0
> 
> Would setting bumpdays to higher values (2 or 3 days) pull in the
> direction I want to go?
> 
> Since bumppercent != 0, bumpsize is ignored from the man page, so I
> guess the only important parameter left is bumppercent. Should I
> increase it or decrease it?  I'm having a hard time understanding what
> the doc says for amanda.conf says...

As you note, your bumpsize is ignored.

The bumpdays parameter gives the minimum number
of dumps at each level that must be done before
the next higher level is even considered.

bumpdays is often recommended to be set to > 1.
The rationale being that if you have two level 1
dumps, one could be lost without much concern.
But if you only have a single level 1, losing it
could be a big problem.

I question your multiplier in "bumpmult  4.0".
It says to go from level 1 to level 2 the level 2
needs to be 20% smaller than the level 1.  But
to go to level 3, the size of the expected level 3
dump must be 80% (4.0 x 20%) smaller than the
expected size of a level 2.  And it will be really
hard to get to level 4 as it must be 320% smaller
than the expected level 3 (4.0 x 80%).

I'd suggest picking a bumpdays with which you are
comfortable, get that multiplier down (1.0? 1.2?),
comment out bumpsize, then see if 20% gives you
what you want, if not start adjusting it.

BTW, if you make a specific dumptype for that
DLE, you can play with the bump settings in that
one dumptype definition without affecting other
DLEs.  They would be using the global values.

Jon
-- 
Jon H. LaBadie j...@jgcomp.com
 11226 South Shore Rd.  (703) 787-0688 (H)
 Reston, VA  20190  (703) 935-6720 (C)


bump* dumptype parameters tuning

2018-04-09 Thread Jean-Francois Malouin
Hi,

It's with some shame that I must admit that through the long years of using
amanda I've never really played with the bump* parameters for DLEs!

I'm now in a situation that might benefit from some tuning using those
to aggressively encourage bumping the gnutar incremental levels.  The
drawback of potentially needing more tapes for recovery is not really a
constraint for me.

I have a DLE (~5TB) for which the estimates are done on the server --
the client, a NAS/filer, doesn't support anything else, so the DLE usage
info is really just from the historical usage not the current one. Any
ideas/hints on how should I proceed from what is currently (default
values I guess) configured?

bumppercent 20
bumpsize20480K
bumpdays1
bumpmult4.0

Would setting bumpdays to higher values (2 or 3 days) pull in the
direction I want to go?

Since bumppercent != 0, bumpsize is ignored from the man page, so I
guess the only important parameter left is bumppercent. Should I
increase it or decrease it?  I'm having a hard time understanding what
the doc says for amanda.conf says...

thanks!
jf