AnsweredAssumed Answered

Red Hat 5.8 webserver node reporting unable to SSH

Question asked by snail on Oct 24, 2012
Latest reply on Oct 28, 2012 by nabeel
I'm running the M5 2.0.1.15869.GA Trial on 64-bit RHEL 5.8 boxes, and it's all working except for the disk tools from a remote host.

As root, I can SSH keylessly between nodes, and the install-mapr script worked fine. If I use the dashboard to get a view of each node's disk status, I get an error very similar to below.

<pre>
2:07:36 PM - Failed to SSH to host node005.
Either the host is unreachable, or passwordless SSH is not set up for user root
Please ensure that the node is reachable and passwordless SSH is configured for user root, and try again.
Alternatively, you may manually run the following commands on the node to list, add and remove disks.
sudo /opt/mapr/bin/maprcli disk list -host 127.0.0.1
sudo /opt/mapr/bin/maprcli disk add -host 127.0.0.1 -disk diskname
sudo /opt/mapr/bin/maprcli disk remove -host 127.0.0.1 -disk diskname
</pre>
For reference, node009 (10.254.0.238) is the web server. node005 (10.254.0.242) is the remote node.

If I SSH to the nodes, keylessly as root of course, I can run the commands.

Sample commands that do work remotely from any node:

<pre>
[root@node009 ~]# /opt/mapr/bin/maprcli disk list -host node005
[root@node009 ~]# /opt/mapr/bin/maprcli disk list -host 10.254.0.242
[root@node009 ~]# /opt/mapr/bin/maprcli disk remove -host 10.254.0.242 -disks /dev/sdb
[root@node009 ~]# /opt/mapr/bin/maprcli disk add -host 10.254.0.242 -disks /dev/sdb
</pre>

Sample commands that do not work remotely from any node:

<pre>
[root@node009 ~]# /opt/mapr/bin/maprcli disk remove -host node005 -disks /dev/sdb
[root@node009 ~]# /opt/mapr/bin/maprcli disk add -host node005 -disks /dev/sdb
</pre>

When coming in remotely -host &lt;hostname&gt;, add and remove don't work. If I check the process list of the node I'm currently shelled into, I can see the SSH, and the remote node reports the SSH login as successful, and I can see a command being run. I have no firewall turned on, and SELinux is disabled. If I do the same command via IP address a the parameter to -host, it works.

Further, these last two commands do not work even if I access the same host I'm logged into, for example, being logged into node005, and providing -host node005 as the option.

An example of this particular failure, trying to remove a disk from node009 itself:

<pre>
[root@node009 ~]# /opt/mapr/bin/maprcli disk remove -host node009 -disks /dev/sdb
ERROR (38) -  Failed to run /opt/mapr/server/diskremove.sh -i 10.254.0.246 -p 7222 -h node009 /dev/sdb
[root@node009 ~]#
</pre>

Use the IP address of node009, and it works correctly.

I'm not sure if it's relevant, following the execution of the command using strace, I see that when using the hostname, for some unusual reason, maprcli tries to connect to 98.227.247.0. This does not occur if I use the hostname for commands that work.

Outcomes