First Time using Rocks (Will this setup work?)

hasuchobe · 11-09-2016, 03:05 PM

Hi all

I have 2 identical computers that I want to do some parallel computations with. It's not an ideal setup by any means. Both computers are connected to a switch which is connected to an ethernet jack by my desk. The head node only has 1 ethernet port (not 2 like the readme describes). I want the head node to also be a compute node. I'm not too sure how to setup the network part but I've been given a hostname and IP address for each computer. Currently I've installed the frontend build on the head node but the DHCPD service won't start (probably cuz I setup the network incorrectly). I think I need the IP address of the switch but I don't know how to get that. Any help would be appreciated!

Ztcoracat · 11-10-2016, 10:46 PM

I've never had to find an ip address of a switch before however;
I found this on my first Google search: "How To get the ip address of a switch"

Quote:

If you want to discover devices in your network try using such tools like NMAP, to map your network. If there is nothing just 2 IP addresses (gateway and IP of PC), so switch has no ip assigned to it (or it is not configured properly). You can use console cable and try to access it.

For DHCP:

Quote:

8.1. Ubuntu client network settings to use DHCP. To configure your client to use a DHCP on a network interface eth0 on Ubuntu or Debian linux systems enter a following lines in your /etc/network/interfaces file: auto eth0 iface eth0 inet dhcp.
8.2. Configuring DHCP relay agent. ...
8.3. BOOTP Support.

http://www.yolinux.com/TUTORIALS/DHCP-Server.html

If your on a server I can not help you; sorry.

Ztcoracat · 11-10-2016, 10:48 PM

Do you mean the Rocks Cluster Distribution?

hasuchobe · 11-21-2016, 05:26 PM

Quote:

Originally Posted by Ztcoracat

Do you mean the Rocks Cluster Distribution?

Yes

I'm actually further along now but I went ahead and bought a NIC for the headnode. The headnode was able to detect the compute node and install whatever it needed but compute0-0 does not show up in Ganglia. Also, running 'tentakel who' in the terminal reports compute0-0 as being down? Not sure what's going on. I basically have the recommended setup for Rocks at this point but compute0-0 doesn't show up after installing. It's very likely I messed up the part regarding eth0 aka the internal network related to the switch.

Also, if anyone can tell me what's the right work flow for using the rocks cluster that would be good too. Like what's the proper way to shut down the cluster when not in use, how do you submit jobs to SGE, how do you setup NFS so that all the nodes have a shared drive. Thanks!

Ztcoracat · 11-21-2016, 06:34 PM

To submit a job to SGE I think you have to write a script. Notice I said 'think'.

http://www.rocksclusters.org/roll-do...atch-jobs.html

http://bioinformatics.mdc-berlin.de/...sing_qsub.html

http://unix.stackexchange.com/questi...chines-san-nas

Good luck-