Dynatrace Agent on JVM with 1 GB heap size doesn't run
We have a JVM instance we're trying to instrument - it runs on a Windows Server (2008?), in concurrence with Tomcat servers. It runs fine on its own, and has an argument assigning a heap size of 1GB to it.
However, when we instrument Dynatrace, it doesn't work unless we assign a smaller heap size to it in the argument.
Line 10 is the original line, which works fine.
Line 8 is what is run when DT is instrumented, and Dynatrace connects, but the script no longer has the memory to run. Does anyone know how we can leave the argument as 1GB, but still get Dynatrace to instrument correctly? What is the problem here?
However, when we instrument Dynatrace, it doesn't work unless we assign a smaller heap size to it in the argument.
Line 10 is the original line, which works fine.
Line 8 is what is run when DT is instrumented, and Dynatrace connects, but the script no longer has the memory to run. Does anyone know how we can leave the argument as 1GB, but still get Dynatrace to instrument correctly? What is the problem here?
hi Randall,
you're using a 32-bit JVM, so I assume the following: windows potentially loads our agent (or some other library) to an arbitrary place in the heap (which it actually can do) so that the JVM cannot allocate the heap as a consecutive block of memory any more and won't start any longer.
we already faced this issue in the past, so we introduced a way to work around this behavior. please try the following:
add ",preallocateheap=true" to the -agentpath option, so that it looks basically like this:
-agentpath:"<path_to_agent>"=name=<agentname>,preallocateheap=true
and try to start the JVM again. this setting basically instructs the agent to preallocate a big block of memory in the very beginning which it then frees again for the JVM to use it for the heap.
let me know if this worked for you.
best,
Christian
