The application placement controller is designed to work with the default settings. However, there can be times when fine tuning the application placement controller becomes necessary to ensure the best results.
Depending on your administrative role, you are allowed the following specific privileges when you are configuring the application placement controller:
Enable | Enables or disables the application placement controller. If you disable the application placement controller, you are disabling all autonomic operations for the dynamic clusters. After the placement controller is disabled, no dynamic changes occur with regard to the size and placement of applications on the dynamic cluster. This action is equivalent to turning all dynamic clusters to static clusters. |
Approval timeout | Controls how long the runtime task sits in the
queue awaiting action before it expires automatically. Acceptable
time values are 1 to 60 minutes. This setting is particular to the
supervised operating mode. When the operating environment is set to
run in supervised mode, the application placement controller creates
tasks, but must wait for approval from the system administrator before
making changes. The application placement controller treats runtime tasks that are timed out as user-rejected tasks. |
Server operation timeout | Represents the amount of time, in minutes, that
the application placement controller waits for a start or stop operation
to complete before the operation is considered a failure. Set this value to the predicted worst case time to start or stop a server. Acceptable values are between 1 to 60 minutes. If a server fails to start before the timeout, the server is placed in a failed server start list. The application placement controller avoids trying to start that server again until you manually fix the start issue and manually start the server. When the server starts, the application placement controller detects that the server has started and removes it from the failed server start list. To get a list of the failed server starts, you can use the PlacementControllerProcs.jacl script. Use the showFailedServerStarts procedure to display a list of failed server starts. See PlacementControllerProcs.jacl script for more information. |
Minimum time between placement change | Specifies the amount of time that the application
placement controller waits before initiating a new batch of changes.
The application placement controller might wait for a batch of changes
after completing previous changes, or encountering a timeout. Acceptable
values can range from 1 minute to 24 hours. When setting this value, consider the overhead that is associated with starting and stopping servers. Starting or stopping servers can take several minutes and might introduce an additional load to the nodes. If you allow the placement controllers to readjust application placement too often, the added overhead negates the increased performance gains that can be earned by readjusting the size of dynamic clusters. For example, if a server takes one minute to start, and the setting for the minimum time between placement changes is 20 minutes, then placement changes have a performance impact of approximately 5%. Set this value at least 20 to 30 times larger than the time necessary to start a server. A value greater than several hours prevents application placement changes from happening more than once a day. If you believe that traffic load and application demands adjust several times during a day, you might want to allow placement changes to occur more frequently. |
The application placement controller is tuned for its best performance.