On Thu, Mar 03, 2011 at 08:22:21PM +0100, Gour wrote:
> On Thu, 03 Mar 2011 10:48:29 -0500
> Jon LaBadie <j...@jgcomp.com> wrote:
>
> > 207KB for each filemark?
> >
> > Wow, what a huge waste!!
>
> Indeed.
>
I meant that with sarcasm, hope it came through.
> >
> > But consider the difference from a 0KB filemark to Amanda's
> > calculations. If it were to write 1000 separate tape files,
> > "wasting" 207MB in filemarks, its calculations of how much
> > dump data it could fit on the tape would differ by 0.1%.
>
> That's right...Still, I wonder whether to try using 0kb for filemark,
> run the test again or something else?
I'd wager you wouldn't be able to differentiate using an LTO-2 tapetype
definition using 0KB vs 207KB filemarks.
Unless things have changed, amanda uses the filemark in calculations
like:
1) during planning it might determine if your DLEs, at their
scheduled backup levels, will fit on the allowed number of
tapes. So unless you only allow 1 tape and each amdump run
gets right up to the 200GB tape capacity, 0.000207 GB per
tape file will have little impact on its plans.
2) while taping, the actual amount of data taped, plus each
32KB tape file header, plus each 207KB file mark, is tracked.
When deciding which completed DLE dump (sitting on the hold-
ing disk) should be taped next, the remaining tape capacity
may be considered depending on the taper algorithim you have
selected (ex the default FIFO vs Largest that Fits). But
those considerations would only come into play if you have
"runtapes" > 1 and are using a changer (even chg-manual).
In my limited experience with various tape formats and tapetype,
I found my LTO-1 drive to give virtually identical results each
run across different individual tapes and 2 brands of tape. The
filemark was always 0. In contrast, my tapetype results for DAT
(DAT-2 and DAT-3) were a bit inconsistant and showed definite
differences between brands.
If my amtapetype results showed a significant filemark while
others generally report zero filemark for LTO formats, I'd ponder,
but wouldn't worry too much, why I got my results. I'd certainly
rerun amtapetype. It not that big a hassle. You could have done
that before asking and had your result before any replies to your
query.
I'd also look at the reported throughput. Amanda makes no use
of the speed results; those are for your info only. I'd compare
my observed results with other reports and with the drive manu-
facturer's specified speed. If your system is unable to feed
your drive data at a rate sufficient to keep it streaming, I'm
pretty sure that apparent capacity and filemark can be affected.
Jon
--
Jon H. LaBadie j...@jgcomp.com
JG Computing
12027 Creekbend Drive (703) 787-0884
Reston, VA 20194 (703) 787-0922 (fax)
