NonStop Server Guide for BEA WebLogic Server 9.2

Configuring Persistent WebLogic Server Processes
HP NonStop Server Guide for BEA WebLogic Server 9.2—529891-002
4-9
Starting a Managed Server Through the Node
Manager
status of a Managed Server as FAILED_MIGRATABLE. In this case, restart the server
using the console or weblogic.Admin utility.
Starting a Managed Server Through the Node Manager
When NativeVersionEnabled is set to false as required on the NonStop server, the
Node Manager uses a shell script to start a Managed Server. By default that script is
named nodemanager.sh but can be changed using the StartTemplate property.
The Node Manager passes four arguments to the script:
•
Java command-line
•
stdout to use
•
stderr to use
•
Filename where the pid of the Java process should be written
Therefore, the nodemanager.sh could be as simple as:
#!/bin/sh
. $WL_HOME/common/bin/commEnv.sh
java $1 > $2 2>$3 &
echo $! > $4
To take advantage of the NonStop server multi-CPU architecture, a sample script and
executable are provided that will balance the Managed Servers throughout the system.
The nodemanager.sh script is stored in $WL_HOME/server/lib/unix. The script
performs some simple processing and then runs a C program, CPUselector, in the
same directory to select which CPU should be used by the Managed Server. The script
then starts the Managed Server.
nodemanager.sh
The nodemanager.sh script reads in a server environment file to allow environment
variables to be set per server. It invokes the C program named CPUselector to
perform the load-balancing functionality. Once CPUselector completes, the script
starts the Managed Server in the appropriate CPU.
The WebLogic Server uses replication groups to support fault-tolerance by allowing a
primary and secondary server to retain state information. To avoid a single point of
failure, do not start the primary and secondary server in the same CPU on a NonStop
server. nodemanager.sh and CPUselector allow you to ensure that the primary
and secondary server do not start on the same CPU.
To support replication groups, CPUs are specified for use using a new environment
variable named WLS_CPUS. Further, the order in which the CPUs are listed in
WLS_CPUS determines the order in which the CPUs are scanned when the load for
each CPU is compared.
nodemanager.sh scans the script arguments to determine the server name to set the
environment variable WLS_SERVER. It then looks for the server environment file in the