Process placement policy and optimization

An active process running only on the master device does not support placement optimization. If you configure a process placement policy for the process, the system displays a configuration failure message. When such an active process fails, the system automatically restarts the process. The standby processes are used for active/standby switchover and ISSU.

Some active processes can run on either the master or subordinate device. When such an active process fails, the system uses a placement policy to select a new active process among standby processes.

The system provides a default process placement policy that takes effect for all processes. You can modify the default placement policy in the view you enter by using the placement program default command. You can also configure a placement policy for a specific process in the view you enter by using the placement program program-name [ instance instance-name ] command. A placement policy for a process takes precedence over the default process placement policy.

By default, the default process placement policy defines the following rules:

• The active process runs on the main CPU of the master, and the standby processes run on the main CPU of the subordinate device.

• A process runs at the location where it ran the last time and does not move to any other location during startup or operation.

• The addition of a new node does not impact current active processes. A new active process selects one node with sufficient CPU and memory resources. (You can use the display cpu-usage and display memory commands to view CPU and memory usage information.)

You can configure the following settings for a process placement policy to optimize process placement:

• affinity location-set—Location affinity, the preference for the process to run on a specific node.

• affinity location-type—Location type affinity, the preference for the process to run on a particular type of node. For more information about node types, see "Configuring a location type affinity."

• affinity program—Process affinity, the preference for the process to run on the same node as a particular process.

• affinity self—Self affinity, the preference for one instance of the process to run on the same node as any other instance of the process.

Affinities include positive affinities (attract) and negative affinities (repulse), all represented by integers in the range of 1 to 100000.

• The higher the attract value, the stronger the preference.

• The higher the repulse value, the weaker the preference.

After you apply new placement policies, the system makes placement decisions based on the new policies, node resources, and topology status. If the new location for an active process is different from the current node, the system changes the state of the process to standby, and uses the standby process on the preferred location as the new active process.