Re: diskless enbd-client/enbd on initrd

"Also sprach dsuchod:"
> more? There are problems with the default cfq scheduler in the kernel, 
> elevator=noop did the thing for me and nbd works without problems in my 
> environment ...

That would disable request aggregation and probably disable plugging. If
so, then requests are treated one by one, at native blocksize, for the
whole kernel. I'm not sure you would like that, but maybe it's better
than nothing!

Mind you, it's not a totally silly idea. I could put in something to do
that on a per-device basis (use ones own make_request fn which does very
little :-).

Peter