Not strictly a cluster question, though thats part of the equation.

I've a BTCPCOM-17 TCP/IP bind() error (49) on a 6.5 cluster node.
Pervasives site KB defines this as:
ID: psql5391
Problem Description:
PVSW.LOG error BTCPCOM-17: A TCP/IP bind () error (49) has occurred.
PVSW.LOG error BTCPCOM-13: A thread initialization error has occured
Communication Modules BTCPCOM and BSPXCOM do not load
Cause of this problem:
Error 49 is described as "EADDRENOTAVAIL (BSDSocket): Cannot assign
requested address. BSTART was set to load in the Autoexec.ncf before
the protocols were loaded

Using any editor Move the BSTART or MGRSTART command to load after the
Mount All command. Restart the server. If it's inconvenient to restart
the server, you may run the MGRSTOP, BSTOP and MGRSTART (or BSTART)
commands at the console and the communication modules will load

The BSTART is ALREADY after the mount all. Through a little digging I
discovered that the (largely undocumented?) BCALLSRV.NLM is loaded as
part of the initsys.ncf stuff, and that in turn loads Btrieve.nlm.

wt# is happening here? Do I need to break out the load of NWMKDE and
of BTCPCOM into separate stages? Nothing I've done within autoexec.ncf
seems to change things, and the other, working-just-fine, cluster node
has the same sequence for all the components discussed.