[syslinux] rpcinfo *very* slow on diskless client

[Adam Nielsen]

Hi all,

Sorry, this might be a bit off topic, but I'm not sure of the best place
to ask.  I'm booting a diskless PC with PXELinux, and the kernel loads
successfully and I can log in and use the system as you'd expect.  I
have done this before, however this time I decided to try using
ClusterNFS so that the diskless clients can share the boot server's
filesystem.

Unfortunately during system boot up to test whether the RPC portmapper
is loaded, the boot scripts run "rpcinfo -p" which causes the PC to sit
there for about ten minutes and then eventually time out and continue
with the boot process.  This happens three times during boot, which
needless to say means there's a *long* wait before I can log in.

Does anyone have any idea why this would happen?  Although the diskless
client and the host share the same filesystem, they have separate copies
of /var/lib/nfs/, /var/lock/ and /var/run/ (among other things) but my
only guess is that there's some other filesystem location that is being
shared and it shouldn't be.

Interestingly enough, the RPC portmapper does eventually load, but none
of the NFS programs work properly, all giving various odd error messages
and seeming to be unable to find the portmapper.  Any ideas?

Thanks,
Adam.