In Java, Metaspace is the memory space used to store class metadata and is separate from the heap. By default, there is no limit, but you can specify a reasonable upper limit to prevent excessive memory use. The limit is set by adding the -XX:MaxMetaspaceSize JVM argument. Use a suitably high limit, such as 512 MB, which is typically sufficient.

This argument is passed from the Core to Worker and it is recommended to keep the values the same in both JVMs.

To set the limit on the Core process, add the argument in the following place and make sure to restart CloverDX Server for the change to take effect:

Java NIO uses direct ByteBuffers for efficient I/O operations, especially when working with native libraries. To improve performance, each thread maintains a small cache of recently used direct buffers in direct memory, known as the temporary buffer cache. By default, this cache has no limit on the size of buffers it can retain, which can lead to excessive off-heap memory usage in multi-threaded applications like CloverDX.

Setting a limit helps reduce memory retention and improve control over off-heap memory use. That is why we, by default, hardcode the limit on the Worker process to 262,144 bytes, which should be sufficient in most environments.

If you need to fine-tune the limit on the Worker process, or add it to the Core process, you can do so by using the -Djdk.nio.maxCachedBufferSize JVM argument. For more information about the argument, refer here.

Add the argument in the following place and make sure to restart CloverDX Server or Worker for the changes to take effect:

Some CloverDX Server installations may experience severe performance slowdowns, sometimes over 100× slower than normal. One possible cause is the JVM code cache becoming full.

The code cache is a special memory area where the JVM stores compiled versions of frequently used parts of the program. When this cache is full, the JVM can no longer compile new code for faster execution and instead must run it in a slower, interpreted mode. This shift can dramatically reduce performance. The reserved code cache size is platform-dependent and can be too small for CloverDX Server.

The limit is controlled by the the -XX:ReservedCodeCacheSize argument. Refer here for more information.

If you hit the code cache limit, you can increase the code cache size by adding the argument on the Core or Worker process in the following places and restarting CloverDX Server or Worker for the changes to take effect: