SAS Micro Analytic Service
is often deployed on a clustered application server to provide high
availability and load balancing. Each node of the cluster can host
one or more instances of SAS Micro Analytic Service as a web application.
Since SAS Micro Analytic
Service uses a GemFire cache for communication between cluster nodes
as well as persistence, there are some implications for scripts to
start up or shut down member instances. Upon starting up a GemFire
cache, cluster members negotiate and determine the member that has
the latest persisted copy of the cache. In order to do this, all members
must be available.
Therefore, any start-up
script must start all cluster members in parallel and
not in sequence. For the Windows platform,
the instances are often created as Windows services, which allows
for parallel start-up. UNIX-based deployments often use a shell script
to restart the instances. Be careful to avoid unneeded dependencies.
If the script starts
the members in sequence (that is, if it starts a member only after
the preceding instance has started successfully), GemFire waits and
delays start-up. Eventually, nodes do start up. However, they might
use an incorrect state of the cache.