Need help with NIS on RedHat 4.2/Alpha

Need help with NIS on RedHat 4.2/Alpha

Post by Neil Henson T-1 » Sat, 15 Nov 1997 04:00:00

Has anyone managed to get an Alpha running RedHat 4.2 set up as an NIS
client correctly? From looking through newsgroup and mailing list
archives, I have learned the following :-

1. There is no built-in support in libc for NYS, so "ybind" does need
to be running.

2. I have installed ypbind-3.0.1 RPM from the RedHat web site and done
the following :-

- created a /etc/yp.conf file containing two lines. One specifies the
NIS domain and the other the ypserver hostname.
- created a /var/yp directory.
- edited /etc/rc.d/init.d/ypbind.init to add a line with the
'domainname' command before the ypbind daemon is started.

Each time I try to log into the machine from across the network, I get
errors in the system log telling me that ypbind has a floating

3. After further searching through the mailing lists, I find that
others report a similar problem and recommend installation of
ypbind-3.3 from :-

The source code compiles out-of-the-box and installs fine. However,
when I run it, I find that the ypbind daemon periodically loses its
connection to the NIS server (the NIS server is a Solaris machine and
is serving the rest of our network just fine during these
disconnects). I also find that it is impossible to ping the NIS server
during these periods (maybe this is a clue to what is going wrong, I
don't know).

I'd greatly appreciate any information from anyone who has had
similiar difficulties or could point me in the direction of some
useful information.



* Neil Henson                          Hartlepool Renaissance: a marina,    *

*        Tel: 505-667-7795  Fax: 505-665-3909  *
* Group T-12, Mail Stop B268, Los Alamos National Laboratory, NM 87545, USA *


Need help with NIS on RedHat 4.2/Alpha

Post by Harvey J. Ste » Thu, 20 Nov 1997 04:00:00

> > Has anyone managed to get an Alpha running RedHat 4.2 set up as an NIS
> > client correctly?


Quote:> I had a similar experience with my Alpha running RedHat 4.2. After 'strace'
> the ypbind slave process, it became clear what was causing the problem.
> Over time, the timeout argument in the select() calls gradually reduced to
> zero. When the value reaches zero, the process blasts the NIS server with
> hundreds of requests until the reply for the first one comes back. I'm
> surprised the server is holding up so well.


Quote:> Could someone shed some light on this?

This is a standard porting to Linux problem.  Linux follows the select
man page suggestion, which says that currently the timeout doesn't
change, but programmers shouldn't rely on this and the timeout should
really be reduced.  So, the problem is that a) Linux reduces the
timeout, and b) programmers ignored the part about not relying on the
timeout not changing.

The fix is that when porting code always make sure the timeout is
reset before calling select.

Good luck,

Harvey J. Stein
Berger Financial Research