clustering with HazelcastStorageService

[Kevin Foote]

> On Apr 27, 2016, at 3:39 PM, Jonathan Johnson <jsjohnson at unicon.net> wrote:
> 
> There are a few deployers out there that are using the Hazelcast storage service in varying levels of deployment. I’ll let them chime in with notes on it if they want. But I can let you know that there are folks using it in production.

Paul, 

Timely query. I just moved to this earlier this week, Tuesday.
Some brief field notes.. :)

The hz session layer has been working fine in our test environment for about 3 months with the latest pull from github. 
Once JJ tracked down our original issue [1] the session storage has not had errors. 

Moving to production the first thing I noticed was the container needed more memory I have it set at 2g currently and 
that seems to be holding. We are replicating same as you idp.session.StorageService, idp.replayCache.StorageService, 
idp.cas.StorageService. 

I have experienced one error that I can not explain at the moment on only a single node. I believe I need to adjust the TCP 
connection parameters a bit, possibly heartbeat if that is possible. 


It would be good to get a page up on sharing conn parameters etc for this thing. (i’ll take that as an action item ;-) 


[1] <https://github.com/UniconLabs/shibboleth-hazelcast-storage-service/issues/2>

--------
thanks
 kevin.foote