WebSphere eXtreme Scale Best Practices for Operation and Management

Chapter 7. Operations and monitoring 143

use, with the total time to start all the containers often being less than if you tried to start

them in groups (even small groups).

For more information, see this website:

http://publib.boulder.ibm.com/infocenter/wxsinfo/v7r1/index.jsp?topic=/com.ibm.

websphere.extremescale.admin.doc/rxsorbproperties.html

3. Start the clients.

When the container servers are active, the grid is available for the clients.

7.1.4 Performing a partial start of container servers

Subsets of the grid can be started in batches of multiple containers by using the WebSphere

Application Server cluster start, or with the WebSphere administrative console. If the

numInitialContainers configuration is being used, placement will not occur until that

threshold is reached, or triggerPlacement on the PlacementServiceMBean is invoked.

You can also use the xsadmin -suspendBalancing and xsadmin -resumeBalancing commands

(or the PlacementServiceMBean) to suspend and resume the balancing mechanism of the

placement service. This action can help reduce CPU utilization when “bouncing” a server in a

topology where there are only a few servers total. If you have two servers and stop one, there

will be no replicas anymore. When you restart the first container on the second server, that

container will be flooded with demands to host replicas. By running xsadmin

-suspendBalancing before stopping the second server and waiting until after all containers on

the second server are restarted before you run xsadmin -resumeBalancing, replicas will be

spread evenly across all the containers. Even if you have more than two servers, suspend

and resume will avoid wasting time replacing replicas multiple times when you are perform a

quick “bounce” of a server.

See 7.3.2, “Useful xsadmin commands” on page 157 for more information about these

commands.

Verifying that a container server start was successful

When start-up occurs, the placement service is responsible for generating a new placement

plan for placing primary and replica shards onto the new containers. To ensure that the

placement service has completed its work for all the new containers, you can check the

placement status to verify that there are no outstanding work items (see “xsadmin

-placementStatus” on page 158).

After the placement service has finished, the xsadmin -containers command retrieves the

catalog server’s view of the current grid placement. Using the same filter parameters that

were described with the teardown command (see “xsadmin -teardown” on page 160), an

operator can get a view of the number of container servers that are running on a host, zone,

or the entire grid.

Example 7-5 Commands to view the current grid placement

xsadmin -containers

xsadmin -containers -fz <zone_name>

xsadmin -containers -fh <host_name>

You can use the current catalog server view of the running topology to confirm that the

number of container servers that you expect to be available matches the catalog server’s

placement view.