Thanks for your advice.
So I will use a solution with includes two steps:
- Set up a clustered GF installation (even if I don't use clustering)
- Set the OnOutOfMemory JVM option in the GF instance
As far as I understand GF clustering, I always need a DAS instance which administrates the cluster. Does this mean, that I need at minimum 2 GF instances - the DAS and one cluster node instance? Or is it possible that the DAS is a cluster node itself where I can deploy my application and be managable by the NodeAgent - I don't think so?
If I really need two GF instances to be able to use the NodeAgent for the "restart on OutOfMemory" szenario, don't you think this is too complex (for a none clustered installation)? Then I have to watch about 3 processes (DAS GF instance, Node GFinstance and NodeAgent process). Also the support stuff have to know exactly in which instance they have to deploy the EAR (in the node GF instance) and that an other GF instance is necessary for administration (the DAS).
Thanks again, greetings,
Frank
[Message sent by forum member 'fmeili' (fmeili)]
http://forums.java.net/jive/thread.jspa?messageID=345958