Automatic Cache Serivce startup only starts Cluster not Hosts while Powershell does - Bug??

Question

Just installed two new servers today (2008 R2) and for the first time installed Appfabric RTM (previously did AppFabric beta2 refresh and Upgraded to RTM and
before that Velocity CTP3)

The following unexpected behaviour occurs when I set the service to automatic startup:

Only the cluster seems to start since only port 22234 is listening in "netstat -an ".

I previously used to call a powershell on startup with the following in it

"powershell -command Import-Module DistributedCacheAdministration, DistributedCacheConfiguration; Use-CacheCluster ;Start-CacheCluster ;"

This now fails to startup the hosts (it says the cluster and hosts are up in fact) and still only port 22234 is listening.

If i revert to how i used AppFabric pre-RTM then all i do keep the service as manual startup and call this powershell.

Then everything works fine and the cluster (22234) and cache (22233) et al are listening and the the cache is usable.

Seems like using auto startup buggers things up and the cache cant be started?

Summary - use autostartup - and the cache wont start (must stop service and start using powershell)

or dont use autostartup and use powershell which is not ideal...

Further info:

Running in a work group - created a user called AppFabric, set the logon as service and generate security audits.

Using XML config:

 <hosts>
            <host replicationPort="22236" arbitrationPort="22235" clusterPort="22234"
                hostId="2070834366" size="3067" leadHost="true" account="ANTLAP\AppFabric"
                cacheHostName="AppFabricCachingService" name="ANTLAP" cachePort="22233" />
        </hosts>

 

 

Answer 1

For starting the services immediately, auto-start is not supported, Start-CacheCluster needs to be invoked. Though services do eventually come up after a few minutes with auto-start. Wait for a few minutes and check the status.

If you are using multiple servers With XML provider auto-start will never work. This is because lead hosts need to come up before the non-lead hosts in a cluster and non-leads go down if they cant find lead hosts. With SQL, you may use auto-start where SQL can act as the lead host, but again it can be a few minutes before the cluster comes up.

Thanks for repoting this, the bug here is the service should show itself as Starting and not Started if it has not come up completely.