You can use the coregroupsplit.py script
to split your existing cell into multiple core groups. Consider running
this script if you have more than 40 WebSphere® Application Server related processes
such as application servers, node agents, and on demand routers defined
in your core group.
Purpose
The coregroupsplit.py script
is available with Version 6.1.0.1 and later.
The
coregroupsplit.py script
divides your existing cell into multiple core groups. If a server
changes its core group membership, then you must restart the entire
cell to prevent partitions from forming. For this reason, this default
script options for script do not change the core group membership
of servers that are members of any core groups other than the DefaultCoreGroup
core group.
This script tries to satisfy the following best practices
for core groups:
- Each core group must have fewer than 40 servers. This best practice
is possible only when you have an adequate ratio of nodes to servers
in your cell.
- Each core group must be equipped with at least three core group
bridges. This best practice is possible only when you have an adequate
ratio of nodes to servers in your cell.
- Each cluster must remain fully mapped to a single core group.
- If you use the -createbridges parameter,
the core group bridges and coordinators are preferably hosted on otherwise
idle nodes.
The coregroupsplit.py script also
tunes the high availability manager for optimum performance. By default,
the script configures a core group bridge on each node agent in the
cell. The script increases the number of core groups until it reaches
the optimal level, depending on the number of nodes and servers in
the cell. The core group bridge node agents are configured as a part
of the DefaultAccessPointGroup access point group in a mesh topology.
In the preferred mesh topology, all of the access points are collected
into a single access point group. As a result, all bridge interfaces
can directly intercommunicate.
Remember that you must give core
group bridges at least 512 MB of Java™ virtual
machine (JVM) space.
Location
The coregroupsplit.py script
is located in the install_root/bin directory.
Usage
The default script usage follows:
./wsadmin.sh -lang jython -f coregroupsplit.py
Running
this script might result in unbalanced core groups, in which some
core groups are loaded more or less than other core groups. You can
rerun this script to rebalance core group membership, but in this
case you must restart the entire cell for the changes to take effect.
To rerun the script, use the following command:
./wsadmin.sh -lang jython -f coregroupsplit.py -reconfig
Parameters
- -reconfig
- Performs a full reconfiguration to rebalance the distribution
of servers among the core groups.
- -createbridges
- Creates separate core group bridge processes instead of creating
the bridge in the node agent.
- -numcoregroups
- Specifies the number of core groups to create.
- -datastacksize
- Specifies a number of megabytes that overrides the default data
stack size.
- -proxycoregroup
- Places the on demand routers (ODR) and proxy servers in a separate
core group.
- -odrcoregroup
- Places the on demand routers (ODR) and proxy servers in a separate
core group.
Example
Use the following example to rebalance
the distribution of servers among all of the core groups:
./wsadmin.sh -lang jython coregroupsplit.py -reconfig
You
can also use this script to create static clusters that are dedicated
as core group bridges for communication within the core group. Use
the following example:
./wsadmin.sh -lang jython coregroupsplit.py -createbridges