Aug 11

Java System App Server first look (part 2)   Posted by sharps, tagged with j2ee java

Clusters, Configs and Node Agents

I found myself with some spare cycles this week so I have been delving into the admin console of the next revision of the Sun Java System Application Server – 2004Q4 (or 8.1 as we call it internally). Today I'm going to write a little about clusters, configurations and node agents – the fundamental concepts necessary to understand the administration model. I'm doing all of this via the admin GUI, not because I can't deal with command lines but because I hope the screens will be self explanatory; I also believe in this case the GUI really does add some value – ie. it really does make some operations much, much easier. Though some of the steps here seem trivial – there is a lot going on behind the scenes. Of course everything I do in the following steps via the GUI I could just as well script / automate via the command line interface (or even the JMX API) – but that is the subject of a future blog.

Task Oriented GUI

If you don't want to go rummaging around the admin console (or aren't familiar with it) there is list of common tasks on the initial screen of the admin console – so you have a more natural entry point for the uninitiated – ie. What do I want to do; not what do I want to do it to.

Create a Cluster

I'm going to create a new cluster, click, ...

I simply give the cluster a name, select a config. (or create a new copy of the default config), add some instances c1-i1 and c1-i2, for each select a node agent (essentially selects the machine the instances will run on). Each machine in the domain will have a node agent – it is basically the local proxy for administration purposes. That's it, I'm done. Though I should note that this is not a 'HA' cluster – ie. I haven't installed or configured HADB – it is essentially a cluster supporting service continuity rather than full session continuity and seamless failover (these terms I will describe in another blog)

I now have a new cluster with 2 instances and a new config. Let's take a closer look at the config (which I can access from anywhere it is referenced or use the navigation tree on the left):

What is a configuration ?

Basically a config is a template used for the configuration of one or more standalone server instances, clustered instances or clusters. Though instances within a cluster would typically be homogeneous; strict enforcement of that would be way too inflexible so we allow the configuration (template) to be parameterized – the screen above shows those parameters. For example, in my case, I'm running everything on the same machine so if strict homogeneity were enforced I would end up with port clashes (as I'm running 5 server instances) – out of the box the wizard knows to parameterize all the port numbers; if I click “Instance Values” I can see the properties (and edit the values) :

So now I have my cluster I need to do something with it (ie. deploy and configure an application) – unfortunately that will have to wait ...

Note : this blog entry refers to a future product therefore the usual caveats apply - everything could change significantly between now and when it is released.