Administer the Cluster Agent

There are several settings available in the Controller for the operation of the Cluster Agent.

Planning for Cluster Agent Deployment

If you have more than one Cluster Agent on your Controller, it is important to start your Cluster Agents in a staggered fashion. This procedure ensures that the Controller responds optimally. Stagger the start time for each cluster as described in the table below. Depending on the size of the cluster if may take a few minutes for the Cluster Agent to fully register with the Controller.

Capacity Planning for Clusters

Capacity planning is important when preparing to deploy the Cluster Agent. The Cluster Agent and the Machine Agent share the capacity of the same Controller. The general rule of thumb is the total number of pods and containers in your cluster is equal to the number of Machine Agents slots that are used. For example, if you have 250 pods and 250 containers, that is the same as using 500 Machine Agents in the Controller. For Controller specifications, see Controller Sizing, and if you have further questions concerning capacity planning, contact AppDynamics Support.

Controller Settings for the Cluster Agent

The following properties are recommended for a Controller supporting the Cluster Agent.

Certain Server Visibility properties will apply to all instances of Server Visibility as well as to the Cluster Agent as indicated below.

Property	Description	Recommended
`sim.cluster.agent.limit`	The maximum number of Cluster Agents per account. Configurable at the account level. For example, If you set the `sim.cluster.agent.limit` to a value of 5, and you have 10 Cluster Agents reporting, the Controller randomly selects a subset of Cluster Agents. These random Cluster Agents will not be registered and reported. If the configured limit is reached at any point in time, any new Cluster Agent will not be registered. See Cluster Agent Requirements and Supported Environments.	10
`sim.cluster.pod.limit`	The maximum number of pods per Cluster Agent per account. Configurable at the account level. If the number of pods exceeds the configured limit at any point in time, the Cluster Agent randomly selects a set of pods that will not register or report. See Cluster Agent Requirements and Supported Environments.	750
`sim.machines.registrations.maxPerSecondPerAccount`	The maximum number of machine registrations allowed per account per second. This property applies to all instances of Server Visibility as well the Cluster Agent.	200
`sim.cluster.container.limit`	Controller level limit for the number of containers per cluster. This will be used if there is no account-level configuration. To overwrite at account-level, add an additional property to the corresponding account. This property only affects the Cluster Agent.	1500
`sim.machines.offline.toStaleTimeoutMillis`	Length of time in milliseconds before an offline machine is considered stale, and hence eligible for purging. This property applies to all instances of Server Visibility as well as the Cluster Agent.	172800000 ms (2 days)
`sim.machines.registrations.update.frequency`	Machine's timestamp update frequency in milliseconds. This property applies to all instances of Server Visibility as well as the Cluster Agent.	86400000 ms (1 day)
`sim.cluster.events.maxPerMinute`	Controller level limit for the number of cluster events that are allowed per minute. This property only affects the Cluster Agent.	4000
`sim.cluster.monitoring.enabled`	This property enables the Cluster Agent on the Controller. If set to false, all Cluster Agents stop reporting metrics and components, but still register every minute. This property only affects the Cluster Agent.	true

See Controller Settings for Server Visibility for further reference.

Note that the sim.machines.offline.toStaleTimeoutMillis value should always be greater than the value of sim.machines.registrations.update.frequency to avoid the risk of purging live entities.

On-Premises

For on-premises customers, follow the steps below to modify the Cluster Agent properties for your account:

Log in to the Administration Console with the administrator root password.
Select Controller Settings.
In the Controller Settings list, search for the setting.
In the Value field enter the recommended value.
Click Save.
Log out of the Administrator Console.

SaaS

Create an AppDynamics Support ticket to have the Cluster Agent properties updated for your account.

Control Cluster Agent Reporting

You can stop and resume the Cluster Agent from reporting metrics and registering containers and pods. This can be done on a per-account basis or on a Controller-wide basis.

Stop a Cluster Agent by Account

To stop the Cluster Agent, on the Controller follow these steps to add the property:

Log in to the Administration Console with the administrator root password.
Select Account Settings.
In the accounts list, double-click the account for which you want to configure the Cluster Agent.
In the Additional Properties screen, click +Add Property.
In the left field enter "sim.cluster.monitoring.enabled".
In the right field enter "false".
Click OK to save the property.
Log out of the Administrator Console.

Start a Cluster Agent by Account

To start the Cluster Agent on the Controller:

Log in to the Administration Console with the administrator root password.
Select Account Settings.
In the accounts list, double-click the account for which you want to configure the Cluster Agent.
Locate the sim.cluster.monitoring.enable property and In the right field enter "true".
Click OK to save the property.
Log out of the Administrator Console.
If the Cluster Agent has been stopped using the "sim.cluster.monitoring.enabled" setting it does not resume reporting metrics and registering pods and containers until you enable the Cluster Agent using the User Interface. See Use the Cluster Agent (GA).

Stop all Agents on the Controller

To stop all Cluster Agents on the Controller:

Log in to the Administration Console with the administrator root password.
Select Controller Settings.
In the Controller Settings list, search for the "sim.cluster.monitoring.enabled" setting.
In the right field set the value to "false".
Click Save.
Log out of the Administrator Console.

Start all Agents on the Controller

To start all Cluster Agents on the Controller follow these steps:

Log in to the Administration Console with the administrator root password.
Select Controller Settings.
In the Controller Settings list, search for the "sim.cluster.monitoring.enabled" setting.
In the right field set the value to "true".
Click Save.
Log out of the Administrator Console.
If Cluster Agents have been stopped using the "sim.cluster.monitoring.enabled" setting they do not resume reporting metrics and registering pods and containers until they are enabled through the User Interface. See Use the Cluster Agent.