After monitoring our servers for the past couple of weeks, I can confirm that the hotfix does not resolve the memory leak issue detailed in my previous post.

Initially, I applied the hotfix to both servers in my cluster and the memory usage continued to rise. However, it was increasing more slowly than before the hotfix was applied.

I re-applied the work around using structClear(variables) in onRequestEnd.cfm and discovered an anomaly. One of the servers was collecting the garbage properly and levelling out at around 200mb usage and the other one wasn't.. The difference was that by mistake one of the servers had -XX:+UseParallelGC included in the JVM config. I applied -XX:+UseParallelGC to both server and they both resumed normal operation with only around 200mb memory usage.

Now, from what I understand -XX:+UseParallelGC is only effective in multi-processor servers, so I couldn't understand why this would make any difference. My guess is that because we're using hyper-threading Xeon processors it's somehow making a difference.

To take the test a step further, I removed the work around again and the memory usage resumed its climb up to around 500mb. However, this time I didn't leave it long enough to hang the server. I'm now running with the work around in place and live in the hope that a permanent fix will become available soon.