Re: ENBD and swap

"Also sprach Dag Sverre Seljebotn:"
> There is currently some activity in the LTSP mailing list regarding
> using stock kernel NBD as the swap mechanism. The tester is experiencing
> some instability though.

He would experience lockups under severe conditions - nothing else. If
he runs under TCP, that is.

> Has ENBD improved upon stock kernel NBD in this area? What issues were
> resolved? Any success stories?

As far as I know, people use it more or less happily, but they are
simply crazy. 

All that enbd does is lock the client code in RAM when you run with
"-s". Then the client cannot be swapped out. After that it's up to
you and the kernel!

Now, if you are using a tcp transport, then tcp will use buffers to deal
with out of order packets, fragmentation, and so on. If RAM is full,
then the kernel must swap something out to get memory for the buffers.
Unfortunately, if swap is over tcp, you are stuck.

Nothing to do with enbd/nbd, in that sense. If you are determined to
use tcp, then you will have to make sure that packets are not
fragmented and not reordered.

You can probably do that by setting blocksize 1KB, not merging
requests, and using a dedicated full-duplex cable connection.


Peter