Managers,
for the second time I've experienced the following problem. I
have two T64 V5.0A boxes (call them host1 and host2) configured
as master and slave NIS servers for our cluster. One of them is
also configured to nfs serve our users' home directories.
This morning I came to work to find that the portmapper on both
boxes had died. Same thing happened also a week ago, still on
monday. Quite obviously all client machines were not exactly
functional. I could determine the portmap on the two machines
above had died by running rpcinfo -p host1 and host2 on an
OpenBSD box whose state was somehow saner (or not so insane)
compared to other clients.
Even more unsettling, though, was the fact that I could not login
as root from console (or via ssh) on either host1 and host2 or
any other Tru64 client! (We have a number of them, all 5.0A and
5.1 --- no patches.) More precisely, the T64 system would let me
(root) in, think about it for a while and then kick me out
without even getting to the shell prompt.
On the other hand, I could log in as root into Linux and OpenBSD
boxes, despite the fact that the Yellow Pages were not working.
All root passwords are local, not served via NIS.
Apart from determining the cause and circumstances under which
the portmaps died, my question to you is: shouldn't the
authentication subsystem return when it finds the root password
in the local database and proceed with the login process without
consulting the yp database? On the contrary, what happened seems
to indicate that somehow, the portmap and/or yp not being
available affects the login procedure anyway.
Thanks, I'll summarize.
--
Ettore Aldrovandi
ettore_at_hep.fsu.edu Ph: (850) 644 3289
http://www.hep.fsu.edu Fax: (850) 644 6735
Received on Mon Feb 12 2001 - 19:10:33 NZDT