MSM and Elasticache node failover

[Erica Kane]

This is a beautiful library. One of things that appeals to me is its resilience in the face of Tomcat or node failure. I think I understand how this works when multiple nodes are involved.

However, AWS Elasticache abstracts its multiple nodes behind one address. If MSM is truly treating this as one node, rather than replicating through the list, I worry that a node failure on Elasticache would result in the loss of session data. (I am using non-sticky sessions.)

One workaround would be to have two separate Elasticache instances, which would cost the same, though this adds complexity of administration. I thought it was at least worth asking, because perhaps MSM *does* store the data in multiple nodes already. And the answer will affect the amount of memory I expect to need.

Thanks,

Erica