Why do my ATG repository item caches end up with usedRatios of well over 100%?

I am running ATG 9 with a bunch of different objects configured in the repository.xml to have specific cache sizes, ttl etc.

For example:

<item-descriptor name="USER" 
    query-expire-timeout="300000"
    item-expire-timeout="300000"
    item-cache-timeout="300000" 
    item-cache-size="20000"
    query-cache-size="50">
...

I am expecting that the cache would not grow above that size and would expire old items to keep the cache size at 开发者_如何学Cor under the item-cache-size. However when I look at the cache stats in the Dynamo admin console, I see several of our items have usedRatios of 500-1000%. This is hogging all of the memory in the JVM over time as more and more items are cached and apparently never released. If I invoke the invalidateCaches method on the Repository in the admin console the free memory jumps back way up and then the slow march down begins again.

How can I ensure that the caches do not grow over their configured size and take over all the memory? Is there some configuration setting I am missing? Are there code tricks one must employ to keep the cache from growing out of control? The ATG docs aren't the most informative and googling around hasn't yielded much info either.

After starting your ATG instance, I suggest navigating to the ProfileAdapterRepository in the Dynamo Admin page (/dyn/admin/nucleus/atg/userprofiling/ProfileAdapterRepository/?propertyName=definitionFiles) and viewing the combined view of the repository definition files. It's the best way to be sure what the final file looks like, since it can be built up of many files.

You should see the attributes you've configured on the "user" repository item through this interface (Note all lower case).

If you don't see your attributes here then you probably don't have your repository definition file loaded, either the module you're working on isn't started or the file is not on the configuration path.