On Thu, May 10, 2007 at 02:40:22PM -0400, Jean-Francois Malouin wrote:
> My first attempt at using tape spanning just failed.
> Anyone can offer some guidance?
>
> Server is an sgi with 28 cpus 28G of ram and
> client is a debian Xeon quad-core with 4GB of ram.
> I've setup an amanda test config for a DLE of ~200GB
> supposed to be stuffed in 100GB LTO1 tapes. The dumptype
> for the DLE reads:
>
> define dumptype tar-span {
> global
> program "GNUTAR"
> priority high
> auth "bsdtcp"
> tape_splitsize 10 Gb
> compress none
> record no
> strategy noinc
> fallback_splitsize 2 Gb
> split_diskbuffer "/holddisk/conf9/split_diskbuffer"
> comment "tape-spanning with tar"
> }
>
> Makes sense or am I using crazy values? Not sure where
> split_diskbuffer lives: on the client or on the server?
> The amdump reports say:
>
client
| via network
V
server holding disk, possibly chunked
|
V
diskbuffer(s)
split1, split2, ... likely one filling, one already filled taping
|
V
tape
> NOTES:
> planner: Adding new disk gaspar:concussion2-span.
> taper: mmap failed (No such device or address): using fallback split size
> of 2097152kb to buffer gaspar:concussion2-span.0 in-memory
> driver: taper pid 13744170 exited with signal 6
>
On my fedora system I had a problem with mmap. It would not handle
buffers of the size I wanted, 2GB. It did finally accept 1GB sized
split_diskbuffers. For me, an OS thing, not an amanda thing.
jl
--
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)
|