I've observed that, when ypbind is used in broadcast mode,
other scripts may be run before the ypbind daemon has
received and processed the response to its broadcast
request.

If amd is also being started at boot time, and expects to
obtain one or more mount maps through NIS, it will
effectively give up on those mount maps if it can't get them
on the first try. In that case, it must be stopped and
started anew -- even a restart via SIGHUP won't help.

The fix I adopted involves adding a loop to
/etc/rc.d/init.d/ypbind, after it has invoked "daemon
ypbind", which tries "ypwhich" once every second until that
program returns a nonzero exit status. (The actual patch
will follow by mail once I have a bug number.)

I've encountered this problem on two different machines
running Red Hat 5.2. A quick inspection of the ypbind script
in 5.9 (Starbuck) suggests that it is vulnerable to the same
thing, but I haven't demonstrated that under conditions of
actual use.

------- Email Received From  Dan Martinez <dfm> 04/01/99 03:59 -------