LinuxQuestions.org - Is it enough to set the IP Address and Hostname in a LAN?

- Linux - Networking (https://www.linuxquestions.org/questions/linux-networking-3/)

- - Is it enough to set the IP Address and Hostname in a LAN? (https://www.linuxquestions.org/questions/linux-networking-3/is-it-enough-to-set-the-ip-address-and-hostname-in-a-lan-599750/)

Is it enough to set the IP Address and Hostname in a LAN?

Hi LQ,

I'm in the process of building a Beowulf cluster. All of the nodes in the cluster are running Linux and are networked together, and I am using the 192.168.0.x address range for my LAN.

I have added the following two lines to each computer's /etc/rc.d/rc.local file, which is a shell script that gets executed after all the other setup happens.

Code:

ifconfig eth0 192.168.0.<number> netmask 255.255.255.0

hostname node<number>

where <number> is the number of the computer.

I've added the necessary aliases to the /etc/hosts files, so that each computer can identify the others by hostname. The nodes use SSH to communicate with each other, and I have set up proper SSH authentication so that no password is required when nodes communicate with each other or run remote commands.

The application I am trying to run on the cluster is returning various TCP errors. Are steps mentioned above enough to effectively change the identities of the computers on the network, or is there more I need to do?

The errors I am getting (if they are any help) are:

Code:

rmcd: getaddrinfo: Temporary failure in name resolution

TCP connect error: Unknown error message.

TCP connect error: return value errno=43

TCP: Connect failed. node01 -> masternode:32809

This is on Aurora Linux, which is Fedora 6 for SPARC chips, running bash.

Thanks very much for your help,
Stephen

Look in your /etc/resolve.conf and /etc/host.conf and /etc/nsswitch.conf files.

Code:

 cat /etc/host.conf

#

# /etc/host.conf - resolver configuration file

#

# Please read the manual page host.conf(5) for more information.

#

#

# The following option is only used by binaries linked against

# libc4 or libc5. This line should be in sync with the "hosts"

# option in /etc/nsswitch.conf.

#

order hosts, bind

Make sure that hosts is listed and it is first.

In /etc/nsswitch.conf, make sure that files is listed first.

Code:

hosts:          files mdns4_minimal [NOTFOUND=return] dns

networks:      files dns

Check that you have the libaries "libnss_files" and "libnss_compat" installed.
Example:
/lib64/libnss_files.so.2

Make sure that you can ping every host by name.

I'm not as familiar with working with clusters. Could the program be trying to set up unix pipe between the hosts instead of going through ssh? Something about the overhead of going through ssh tunnels in a cluster seems counter productive to me, but you have far more experience with clusters than I do ( which for me is none ). Do you have ssh tunnels set up for certain ports that this program uses?