Ask Questions and Find Answers

Really, that link you keep sharing is total crap...

First, we don't need fake entities in 7, it is easier to build API and Impl modules w/o the SB overhead.

But more importantly, what he documents isn't a cache as much as it is an "in memory data sink". Caching implies expiry, flushing, size limits, etc. That implementation you're pointing to is not a cache.

We don't know what you're trying to do, so it is hard to recommend anything.

The only thing you posted was that you tried implementing something from a blog somewhere that is untried, nonstandard and unnecessary.

So if you could start over and explain what you needed, we'd be happy to offer the right path to accomplish what you need...

So I would actually recommend leveraging SingleVMPool for this. It handles storing data in ehcache for you, so you get cache sizing, expiration, etc but you don't have to worry about doing manual cache management yourself.

I found this reference that can help you leverage it: http://wwwilpower.blogspot.com/2015/01/adding-custom-cache-to-liferay.html

I would not suggest using MultiVMPool for this. MultiVMPool broadcasts across the cluster, so nodes could be loaded down with data they don't need just because another node has fetched the info. Let each node use its own cache so it holds data it needs to service requests that it gets.

As far as implementation goes, it would be a simple matter of asking SimpleVMPool to return your cached value; if you get one, return it, if not it is a cache miss and you should hit your web service. Remember to put the web service result into SimpleVMPool before returning it so it is in the cache for the next value.

This logic would go right in where you are currently just invoking the web service directly now, so it is easy to fit in to what you already have.