[ENBD] Getting nbd (2.4.26) to work!?

[Peter T. Breuer]

"A month of sundays ago Johann Spies wrote:"
> We are in the process of setting up a ftp-server which will utilize 4
> computers: the server and three older computers with 3x40G IDE hard
> disks on each linked with a 100Mb network cable to the server.  We
> were thinking of using nbd to mount the hard disks on the server, but
> are beginning to doubt whether nbd will be the right tool for the job
> after trying for a few days now to get a simple nbd connection to
> work.
> 
> The following output (on the client side) shows our problems as it is
> on the moment (The server is presenting a file just less than 2G on
> which mke2fs has made a file system on port 4017):
> --------------------------
> # nbd-client: client (-1) starts introduction sequence on port 4017
> stream: connect_sock client socket connect failed Connection refused
> on port 4017
> nbd-client: introduction Connect failed: Connection refused

This tells you there is no server.

Give up now, and fix the server. It isn't there.

>  mount: wrong fs type, bad option, bad superblock on /dev/nda,
>         or too many mounted file systems

Perfectly correct and consistent.

> The 2.2.26 was compiled and installed on both.

Surely you mean enbd 2.4.26?

> Now my questions:
> 
> 1. Can nbd be used for the purpose we intended as described above?

I don't know. I didn't understand your explanation of what you were
trying to do - to be honest, I didn't see anything labelled as
an explanation. The paragraph I've left intact at the head of this reply 
is about as much as I saw. It says you want to bring 3x 3x40GB across
to a central machine via nbd. This is perfectly normal. But I don't
knwo the details. How are you proposing to aggregate these, both
on the server side and on the client side?

If I were you I would aggregate them via raid on the server side,
export as a single 120GB partition, then aggregate again via raid on
the client side, as a single 360GB partition (or whatever).

> 2. How do I get it to work?  Every time we want to do something like
>    mke2fs on /dev/nda or try to mount it, the client died.

No it didn't. The server died. Or isn't there!

>        Aggregated resources may together exceed the maximum file
>        system size limit (cur_ rently 2GB on linux intel) so long as
>        their components are each within the limit.
> 
>    Does that imply that one can not mount a partition larger than 2G?

This is from the server documentation, surely?

No, it says and implies the opposite - even if your system is limited
to 32 bit file sizes, you can export aggregates of more than 2GB total
via the server's built-in raid-linear facilty.

Provided it is working nowadays, that is! I doubt it's been tested
recently. There is no need to use it, since your distribution probably
has a 64bit llseek, and your kernel certainly does.

Let me see the server command line.

Peter