I have a distributed cluster of close to 100 vms at the moment. Its quite the challenge to debug when something goes wrong so we are writing a distributed query to query into the caches for the developers to debug the contents of the cache cluster.
The challenge is we have a myriad of caches with different distribution and deployment profiles over the cluster. This is needed or every instance would be a minimum of 6Gb in memory footprint just to hold the class definitions of our backend. Ofcourse this creates a small challenge when it comes to unmarshaling
Our plan was to do a distributed query and translate the object on the local cache instance (since it knows the class definition) to json and return this to the dev query app so we easier can debug this.
My hope was to find a localised method like keySet, size and so on where I can used the distributed execution framework to do a local query on all the caches, transform it to json and return it to the calling point where I can do map reduce and present json entries for the developers so they feel connected with the cache and can debug what is going on inside there easier.
I would be very happy for any tips on how to solve this (monitoring is out of the question as our ASP vs firewall setup is a pain to get through when it comes to JMX and the sort.)