If you have read part 1, then you will know that I left the installation partly incomplete (poorly configured). To resolve the issues I raised last time, a secondary disk was added to the server I nominated as the Head Node, and all the Network connections were named correctly.
I now have 4 full working and connected nodes; each node was installed manually by using the “Add Node (wizard)” which is available from the MMC on the Head Node. When you have the nodes installed, there are four main actions available for administering the nodes.1. “Launch a Remote Desktop Connection” – does just what it says, starts a RDC connection which is displayed in the central MMC (v2.1/3.0) viewing area.2. “Execute a command” – more of this below…3. “Identify a node” – just pops out the CD-ROM drive (so you can find the server in the rack)4. “Open System Monitor” – Opens a system monitor (per node) that is pre-populated with some useful Perfmon counters
I left “execute a command” as I wanted to explain in a bit more detail what it does by giving some examples. To use this feature you highlight a node, or use CTRL or SHIFT to select multiple servers and then right click and select “Execute a Command”. You then get a simple dialog box, the first part is drop down text box and a results window. In the text box you can type the command you want, for instance I wanted to confirm the IP configuration for all my nodes, so I typed IPCONFIG into the box, and pressed the execute button. In the results window the 4 nodes returned the data (each server is separated with it name in bold in the returned list, so it’s easily readable when you scroll down). The next task I needed to do was to create a shared directory on all my nodes, so I just typed: cmd /c MD C:\DEMO and hit execute to create the directory (note: the /C switch tells the cmd.exe to run then close when finished), I then typed: cmd /C NET SHARE demo=c:\demo which shared out those newly created folders. I then needed to copy some file to these shares, which was easily done by typing: cmd /C copy \\headnode\source\*.* c:\demo\*.* which copied the source files from my head node server (called headnode) to the c:\demo directory on all the nodes. The whole process I think is much easier than typing these commands on all the servers – do you agree?
I should have also said when a new node is first added, either by the “Add Node wizard” or added manually from the node, then you will need to go to the “Compute Cluster Administrator” thats what the MMC is called, and “approve” the new node, as well as performing a “Resume”. The approves process basically means that an administrator has to conform that the node is to be added to the cluster, the resume take the node out of the “pause” state, which means it is allowed to runs jobs. You can “pause” a node at any time if you require, which means that no new jobs will be accepted, but existing jobs will run until completion.