All-
Openmpi1.3.2 is the current version running on our cluster.
Our authentication method is Kerberos and /usr/kerberos/bin/rsh is used for transport.
To forward a Kerberos5 ticket to another node we use regularly "rsh -F"
dqcneh001$ rsh -F  dqcneh002 klist
Ticket cache: FILE:/tmp/krb5cc_p3008
Default principal: jdu...@crc.nd.edu

Valid starting     Expires            Service principal
08/04/09 17:02:33  09/02/09 22:37:22  krbtgt/crc.nd....@crc.nd.edu
        renew until 09/02/09 22:37:22
08/04/09 17:02:33  09/02/09 22:37:22  afs/crc.nd....@crc.nd.edu
        renew until 09/02/09 22:37:22


Kerberos 4 ticket cache: /tmp/tkt82784

The logs on dqcneh002 shows the remote connection and the spawn of kshell
---------
Aug 4 17:02:33 dqcneh002 xinetd[3095]: START: kshell pid=3008 from=10.32.10.68
Aug  4 17:02:33 dqcneh002 kshd[3008]: Shell process completed.
Aug 4 17:02:33 dqcneh002 xinetd[3095]: EXIT: kshell status=0 pid=3008 duration=0(sec)
----------


When I try
dqcneh001$ mpirun -np 1 -H dqcneh002 -mca plm_rsh_agent "/usr/kerberos/bin/rsh -F" klist
klist: No credentials cache found (ticket cache FILE:/tmp/krb5cc_p3651)


Kerberos 4 ticket cache: /tmp/tkt82784
klist: You have no tickets cached

The Kerberos ticket is not forwarded.

The logs on dqcneh002 shows that kshd is sstarted properly though:
Aug 4 17:13:32 dqcneh002 xinetd[3095]: START: kshell pid=3651 from=10.32.10.68
Aug  4 17:13:32 dqcneh002 kshd[3651]: Shell process completed.
Aug 4 17:13:32 dqcneh002 xinetd[3095]: EXIT: kshell status=0 pid=3651 duration=0(sec)


Am I missing something?
Thanks for any help
JC

Reply via email to