[ENBD] Client freeze with Reiserfs over enbd

[Peter T. Breuer]

"Brian Keefer wrote:"
> "Peter T. Breuer" wrote:
> > If you have the kernel debugger compiled in, you can generally get out
> > of jail (blind). Maybe with magic-sysreq too.
> 
> With magic sysreq, moya couldn't sync or unmount it's disk. I could only
> Sysrq-b, and offer my humble thanks to Hans Reiser, Patron Saint of Fast
> Booting.

Ah - but now I understand. The journal ws on the same device as you
had the filesystem on. That's the problem. It'll produce i/o storms
over the network, to say nothing of not actually protecting you
from anything that I can think of like that.  I think that
by default write order is not being preserved (you need -w something)
which means that the journal is not necessarily being written in the
right order either.

> > Isn't that interesting. The reply came in, but 13s late!
> > 
> > What could do that?
> 
> Umm, well, the Server here is supposed to be a diskless mediastation,
> shangai'd into being a cheap external IDE enclosure. The Server's
> enbd-server binary is being Samba'd in from moya. I know Linux's IO sorter
> sucks, but it wasn't too busy to allow samba to reply to enbd-server being
> paged in. 

I think the -s switches make the binaries always paged in, if you have
the ram.

Peter