Alfresco Cluster - L2 Cache

[pedrocerdeira]

I'm a web web developer with some experience in Java and .Net for large public scale systems.
I'm currently looking into Alfresco (latest version) in order to evaluate it's future use in future projects.

Disclaimer: i do not want to focus enterprise or community.

One of the topics i have to evaluate is clustering. I want to focus on Alfresco.war clustering for now.
From what i understand (http://wiki.alfresco.com/wiki/Cluster_Configuration_V2.1.3_and_Later#High_availability_components) in order to cluster Alfresco.War, i need to sync L2 cache for each server. Current enterprise version uses hazelcast to achieve that.

Two questions:
-(1) Why is it important to share L2 cache? Is it for performace issues only, or if i don't it could malfunction? (I would like to understand from a dev point-of-view)
-(2) It is possible to disable L2 cache. With L2 cache disabled, will i be able to cluster without any configuration?

With this i would like to understand why do i need to share l2 cache.

Thanks,
PC

[jspuchau]

Hi Pedro,

Before answering you telling taht from what I now, clustering capabilities are only available on Enterprise Editions.

On the other hand, I don't know if it's possible to disconnect the L2 cache, but you can avoid to share it. Just don't update the ehcache config file. Anyway, you'll have problems on consistency, as the indexes are not going to be replicated between the nodes in the cluster.

So, imagine you upload a content through node A, node A indexes are updated with the info of the content, but when you try to perform a search related to the uploaded content on node B, no contents are going to be shown to you, as node B indexes are not updated.


Surely, other problems will appear, but I think this is the most important.

Regards

[ffaizan]

Is it possible to use the same lucene index and content store across both the servers in the cluster.
I am not sure if it works but if it does than there should not be a need for cache replication as both the
servers should be using the same lucene index.

[mrogers]

The bottom line is that clustering is enterprise only.   It's more than just the L2 cache.

You shouldn't need to worry about the implementation details.