This page describes the resource requirements, compatibility with different Kubernetes-based software, Cluster Agent distribution, licensing, and performance specifications.

Hardware Requirements

The default requirement for the hardware is as following:

  • CPU: 1250m
  • Memory: 300Mi

These values are applicable when you use Cluster Agent without auto-instrumentation. The values can change when you perform auto-instrumentation.
With auto-instrumentation, you need to increase the memory limits by 50Mi for every 100 pods. For Example, If 500 pods are instrumented through the agent, add 550Mi (300Mi + 250Mi) to the memory limits value.
You need to increase the CPU limits by 100m (1350m) when auto-instrumentation is enabled.

Software Requirements

The Cluster Agent is designed to run on Linux and deployed using the AppDynamics Operator.

The Cluster Agent requires:

  • AppDynamics Controller >= 20.3.0.
  • Kubernetes versions, 1.11, 1.13, or >=1.14, with the Kubernetes metrics-server deployed and enabled on the cluster.
  • A cluster that you can access and monitor.
  • Sufficient Server Visibility licenses. The Cluster Agent consumes one Server Visibility licenseSee License Management

  • Access to Docker Hub or Red Hat Container Registry to pull the Cluster Agent Operator and Cluster Agent images, or access to an internal repository where these images are maintained. See Install the Cluster Agent.

If you deploy the Cluster Agent to monitor your Kubernetes Cluster, it does not monitor worker nodes. To monitor worker nodes, you must install the Machine Agent which consumes additional Server Visibility licenses (one Machine Agent per node). See Install Infrastructure Visibility with the Cluster Agent Operator

The Cluster Agent is compatible with these cloud platforms:

Cloud PlatformVersion
Rancher Kubernetes Engine (RKE)
  • 1.0.4 with Kubernetes 1.17
  • 1.2.1 with Kubernetes 1.19.3
Kubernetes1.11, 1.13, 1.14, 1.17, 1.19, 1.20, and 1.21 with the Kubernetes metrics-server deployed
Amazon EKS1.11, 1.13, 1.14, 1.15, 1.16, 1.17, 1.18, 1.19, and 1.20 with the Kubernetes metrics-server deployed
Azure AKS 1.11, 1.13, 1.14, 1.15, 1.16, 1.17, 1.18, 1.19, 1.20, and 1.21 with the Kubernetes metrics-server deployed
Google GKE 1.14, 1.15, 1.16, 1.17, 1.18, 1.19, and 1.20
Red Hat OpenShift

3.11, 4.1, 4.2, 4.3, 4.4, 4.5, 4.6 and 4.7

The metrics-server is not shipped with OpenShift, you must deploy it separately. The CRI-O container runtime is supported on >=4.2.7.

kubectl>=1.11.3


Installing AppDynamics to test or evaluate involves verifying system requirements, preparing the host, and then installing the Controller. See Install the Controller Using the Enterprise Console and Install the Controller Using the CLI.

Cluster Agent Distribution

You can download the Cluster Agent from the AppDynamics Download site as a zip archive. The archive also contains distribution-specific configuration files for the deployment.

Licensing

The Cluster Agent requires a Server Visibility license. To get started, see Enable Server VisibilityIf you already have a Server Visibility license, see Install the Cluster Agent for installing Cluster Agent on a Kubernetes cluster.

Cluster Agent Performance Certification

 

Cluster Agent VersionPerformance Certification
<=20.10

You can monitor the certified stable limit of 750 pods and 1500 containers

>=20.11
  • Cluster Agent Vertical Scaling:
    This is the certified stable limit:
    If you run one container per pod then you can monitor 2250 pods and 2250 containers.
    If you run two containers per pod, then you can monitor 1500 pods and 3000 containers.

Cluster Agent and Pod Limits

On-Premises

You can configure the number of Cluster Agents based on the overall controller capacity. To configure the limit, see the sim.cluster.agent.limit description under Controller Settings for the Cluster Agent

If your Cluster has more pods than can be monitored, the Cluster Agent ensures that the group of pods being monitored remains the same over time. If any of the reported pods are terminated in the future, then the Cluster Agent replaces the terminated pods with previously unreported pods and starts monitoring the newly included pods.
The total pods metric displayed on the Pods List page reflects only the number of pods up to the configured limit. See the sim.cluster.pod.limit description under Controller Settings for the Cluster Agent.

SaaS

To configure the limits, create an AppDynamics Support ticket.