Resolving The Problem

When running "pdsh -a uptime", I received errors from all the cluster hosts similar to:

[root@installer ~]# pdsh -a uptime

compute-01-00-eth0: Warning: Permanently added 'compute-01-00-eth0' (RSA) to the list of known hosts.

compute-01-00-eth0:

compute-01-00-eth0: (gnome-ssh-askpass:24971): Gtk-WARNING **: cannot open display:

compute-01-00-eth0: Permission denied, please try again.

compute-01-00-eth0:

compute-01-00-eth0: (gnome-ssh-askpass:24972): Gtk-WARNING **: cannot open display:

compute-01-00-eth0: Permission denied, please try again.

compute-01-00-eth0:

compute-01-00-eth0: (gnome-ssh-askpass:24973): Gtk-WARNING **: cannot open display:

compute-01-00-eth0: Permission denied (publickey,gssapi-with-mic,password).

pdsh@hpc201d: compute-01-00-eth0: ssh exited with exit code 255

The passwordless ssh between the nodes in a PCM cluster is configured by default behavior.

When a PCM installer node is installed, in addition to the authorized_keys file of the installer located under /root/.ssh, an authorized_keys file located under /opt/kusu/etc/.ssh/ is generated to be distributed to all the managed compute nodes only.

In some case that user updated root's ssh key pair on the installer node, the passwordless ssh will be broken.

To resolve this issue, on the installer node, do the following as root user:

# cat /root/.ssh/id_rsa.pub >> /opt/kusu/etc/.ssh/authorized_keys

Then, run:

# cfmsync -u