testing clustered opendb

[tridge at samba.org]

re-sending as I speled samba-technical wrung the frst tyme

-----

Ronnie and Volker,

I've been thinking a bit more about how to properly test clustered
opendb. I think the key will be to do more validation of what files
should be allowed to be opened in the torture client.

So for example if we modified RAW-BENCH-OPEN to open more than one
file at a time on each connection, and kept a table inside the test of
what connections have what file open (we can just re-use the state[]
array for that), then we could flag an error if the same file was ever
open on two connections at once.

This might work best with a change to make the test choose random
files to open, rather than the round robin it currently does.

We need to be a bit careful with the async nature of the test. We only
know for sure that we have the file open on a connection when we have
received an async open reply, and haven't yet sent the async close
request. That's why we need to open more than one file at a time per
connection - otherwise we'd never know for certain that two
connections have the same file open (currently it sends a close as
soon as the condition is met).

Cheers, Tridge