On Wed, Apr 30, 2003 at 10:33:33AM -0400, Eric Sproul wrote:
> Hi,
> I have a question about how the planner schedules dumps and its
> interaction with the tapetype definition.
>
> In my home config, I use a Seagate Scorpion 24 with DDS2 tapes and the
> following definition:
>
> define tapetype Python-DDS2 {
> comment "ARCHIVE Python with DDS2 tape"
> length 3860 mbytes
> filemark 16 kbytes
> speed 707 kps
> }
>
> 3860 MB/1024 = 3952640 KB
>
> I think this is sensible, knowing that DDS2 is 4GB native (4194304 KB).
> I use software compression, so AMANDA should have an accurate idea of
> how much data she wants to write. From the tapetype above, she should
> know how much tape she has to play with, and I should hope that even if
> she writes all 3860 MB, that still leaves a little cushion at the end of
> the tape.
>
> Trouble is, I've seen AMANDA try to write more than this to a tape, and
> hit an EOT. She tried to write more data than the tapetype told her she
> could. That's what I don't understand.
>
> For example:
>
> STATISTICS:
> Total Full Daily
> -------- -------- --------
> Estimate Time (hrs:min) 0:09
> Run Time (hrs:min) 1:50
> Dump Time (hrs:min) 1:31 1:29 0:01
> Output Size (meg) 3846.6 3844.9 1.7
> Original Size (meg) 4023.8 4002.5 21.3
> Avg Compressed Size (%) 60.2 62.8 7.8 (level:#disks
> ...)
> Filesystems Dumped 15 11 4 (1:4)
> Avg Dump Rate (k/s) 722.2 734.0 18.9
>
> Tape Time (hrs:min) 0:59 0:59 0:00
> Tape Size (meg) 2450.8 2449.0 1.8
> Tape Used (%) 63.5 63.5 0.0 (level:#disks
> ...)
> Filesystems Taped 14 10 4 (1:4)
> Avg Tp Write Rate (k/s) 705.5 705.7 546.3
>
> and in NOTES:
>
> taper: tape small07 kb 3896224 fm 15 writing file: No space left on
> device
>
> So it tells me that when AMANDA hit that EOT, she was already beyond the
> limit defined in the tapetype. Why was she doing that? I'm interested
> to know if I need to tweak the tapetype a little more.
That was your last DLE to go to tape.
I presume you have runtapes set to 1, not multiple tapes per session.
Amanda's choices as I see it were:
1) it doesn't look like it will fit,
report that and leave it on the holding disk
2) it doesn't look like it will fit,
but it doesn't hurt to try anyway and if it fails,
report that and leave it on the holding disk
Note the last line of each choice is the same.
I have runtapes set to multiple tapes and noted a similar thing. Amanda
would try to squeeze a dump onto the tape that clearly (to me anyway)
would not fit. Yet there were other dumps avaliable that would fit.
Seemed like a waste of good tape. I had just switched to 2.4.4 and
some kind sole (thanks JRJ) pointed out I had not RTM and noticed
the new 'taperalgo' parameter. Don't think it would help your case
with runtapes set to 1 and no other DLE's.
> On a related note: Does anyone know whether native tape capacities are
> advertised the way hard drives are (where 1GB = 1000 MB instead of 1024
> MB). Although even if this were the case, I'd still expect to get
> 4096000 KB on the tape, and AMANDA hit EOT at 3896224 KB, almost 200 KB
> shy. How much variance is there in actual usable capacity vs.
> advertised native capacity?
I always believed marketing used 1G = 1,000,000,000. That would make
a dds2 tape 3.725 of 'our' G's. Observed results seem somewhere in between.
Way back, I used to save each tapetype published on this list. Collected
a quarter meg of email by the time I stopped. I don't recall a single
one ever reporting that tapetype found a capacity greater than the published
native capacity.
--
Jon H. LaBadie jon AT jgcomp DOT com
JG Computing
4455 Province Line Road (609) 252-0159
Princeton, NJ 08540-4322 (609) 683-7220 (fax)
|
