AppDynamics Operator simplifies the configuration and lifecycle management of the AppDynamics ClusterAgent and the AppDynamics Machine Agent on different Kubernetes distributions and OpenShift. The Operator encapsulates key operational knowledge on how to configure and upgrade the ClusterAgent and the Machine Agent. It knows, for example, which configuration changes are benign and do not require restart of the ClusterAgent, which minimizes unnecesary load on the cluster API server.
The Operator is implemented using OperatorSDK and uses Kubernetes API to maintain the desired state of the custom resources that represent the ClusterAgent and the Machine Agent. When the Operator is deployed, it creates custom resource definitions (CRDs) for 2 custom resources:
- clusteragent, which represents the ClusterAgent.
- infraviz, which represents the Machine Agent bundled with netviz and analytics.
This level of abstraction further simplifies the management of monitoring and imstrumentation and ensures granular security policy of the ClusterAgent and the Machine Agent.
Create namespace for the operator and the ClusterAgent
-
Create namespace for AppDynamics components
- Kubernetes
kubectl create namespace appdynamics - OpenShift
oc new-project appdynamics --description="AppDynamics Infrastructure"
- Kubernetes
-
Create Secret
cluster-agent-secretwith the following key- The "controller-key" - the access key to the AppDynamics controller.
kubectl -n appdynamics create secret generic cluster-agent-secret \
--from-literal=controller-key="<controller-access-key>" \
- Update the image reference in the Operator deployment spec (deploy/cluster-agent-operator.yaml), if necessary.
The default is "docker.io/appdynamics/cluster-agent-operator:latest".
- Deploy the Operator
kubectl apply -f deploy/cluster-agent-operator.yaml
By default "docker.io/appdynamics/cluster-agent-operator:latest" is used.
AppDynamics images are also available from Red Hat Container Catalog.
To enable pulling, create a secret in the ClusterAgent namespace. In this example, namespace appdynamics is used and appdynamics-operator account is linked to the secret.
$ oc -n appdynamics create secret docker-registry redhat-connect
--docker-server=registry.connect.redhat.com
--docker-username=REDHAT_CONNECT_USERNAME
--docker-password=REDHAT_CONNECT_PASSWORD --docker-email=unused
$ oc -n appdynamics secrets link appdynamics-operator redhat-connect
--for=pull
The AppDynamics Cluster Agent is a new, lightweight agent written in Golang providing APM monitoring for Kubernetes. The Cluster Agent is designed to help you understand how Kubernetes infrastructure affects your applications and business performance. With the AppDynamics Cluster Agent, you can collect metadata, metrics, and events about a Kubernetes cluster. You can query for the components in the cluster, and report metrics for those components. The Cluster Agent works across cloud platforms such as Kubernetes on AWS EKS, AKS on Azure, PKS on Pivotal and many others.
The clusteragent is the custom resource that the Operator works with to deploy an instance of the ClusterAgent. When a clusteragent spec is provided, the Operator will create a single replica deployment and the necessary additional resources (a configMap and a service) to support the ClusterAgent.
Here is an example of a minimalistic spec of the ClusterAgent custom resource:
apiVersion: appdynamics.com/v1alpha1
kind: Clusteragent
metadata:
name: k8s-cluster-agent
namespace: appdynamics
spec:
appName: "<app-name>"
controllerUrl: "<protocol>://<appdynamics-controller-host>:<port>"
account: appdynamics-cluster-agent
image: "<your-docker-registry>/appdynamics/cluster-agent:tag"
Update the provided spec with the AppDynamics account information and deploy:
kubectl apply -f deploy/cluster-agent.yaml
Link to official Cluster Agent Installation Guide
Appdynamics operator can be used to enable server and network visibility with AppDynamics Machine agent.
The operator works with custom resource infraviz to deploy the AppDynamics Machine Agent daemon set.
Here is an example of a minimalistic infraviz spec with the required parameters:
apiVersion: appdynamics.com/v1alpha1
kind: InfraViz
metadata:
name: appd-infraviz
namespace: appdynamics
spec:
controllerUrl: "https://appd-controller.com"
image: "docker.io/appdynamics/machine-agent-analytics:latest"
account: "<your-account-name>"
globalAccount: "<your-global-account-name"
The controller URL must be in the following format:
<protocol>://<controller-domain>:<port>
Use the provided specs for Kubernetes and OpenShift to deploy the Machine Agent. Make sure to update the spec with the AppDynamics account information prior to deployment.
kubectl apply -f deploy/infraviz.yaml
On OpenShift:
oc apply -f deploy/infraviz-openshift.yaml
As of v 0.5.0, the operator can deploy Machine Agent daemonsets to both Linux and Windows nodes. The deployment startegy in mixed OS clusters is determined by the value of the nodeOS property. This property has the following values:
- linux
- windows
- all
To deploy the Machine Agent to both OSs you have several options:
-
A single Infraviz resource. You may use a single custom resource of the type Infraviz to deploy the Machine Agent across all nodes regardless of the operating system. Set the
nodeOSproperty toall. You also have to provide the windows image reference in theimageWinproperty of the InfraViz spec. When thenodeOSproperty is set toall, the operator will create 2 daemonsets, one for Linux nodes and the other one for Windows nodes. -
Separate Infraviz resources. You can create 2 independent custom resources of the type Infraviz, one for the Linux nodes and the other one for the Windows nodes. In the Linux spec set the value of
nodeOStolinux. In the Windows spec, set the value of thenodeOStowindows. In the Windows spec you also have to provide the image reference in theimageWinproperty. Make sure that the custom resource names are different.
The choice of the strategy depends on your customization needs. The first strategy is convenient when both Linux and Windows daemonsets share all Infraviz properties (resources, ports, etc). If Linux and Windows daemonsets need to be customized independently, the secod strategy is more practical.
In both secanrios the placement is driven by the nodeSelector value. The operator will generate a nodeSelector, if not specified by the user, and set the value of "kubernetes.io/os" to linux or windows depending on the value of the nodeOS property. You can provide additional nodeSelector values as necessary using the nodeSelector property of the Infraviz spec.
If nodeOS is not set, the operator will attempt to deploy the daemonset to all available worker nodes in the cluster, regardless of the OS. If you have a mixed OS cluster, you will need to set the nodeSelector property to kubernetes.io/os: linux or a similar depending on the labels used on the cluster nodes.
The enableMasters directive will be honored in all cases except when nodeOS property is set to windows.
Updates to the infraViz with different nodeOS values is transparent. For example, if you have an existing Infraviz instance with nodeOS set to "", "linux", or "windows" and you change the nodeOS property to 'all', the operator will automatically create the second daemonset and update the existing one if required. There is one excpetion, when going on reverse, from 2 daemonsets down to 1 daemonset. In this case you need to delete the Infraviz resource with nodeOS set to all and deploy another Infraviz instance targeted for Linux or Windows
| Parameter | Description | Default |
|---|---|---|
controllerUrl |
Url of the AppDynamics controller | Required |
account |
AppDynamics Account Name | Required |
globalAccount |
Global Account Name | Required |
eventServiceUrl |
Event Service Endpoint | Optional |
enableContainerHostId |
Flag that determines how container names are derived (pod name vs container id) | "true" |
enableServerViz |
Enable Server Visibility | "true" |
enableDockerViz |
Enable Docker Container Visibiltiy | "true" |
uniqueHostId |
Unique host ID in AppDynamics. | Optional. If not provided, the operator will set the value to spec.nodeName. Valid options: spec.nodeName, status.hostIP |
metricsLimit |
Number of metrics that the Machine Agent is allowed to post to the controller | Optional |
logLevel |
Logging level (info or debug) |
info |
stdoutLogging |
Determines if the logs are saved to a file or redirected to the console | "false" |
syslogPort |
The embedded analytics agent uses this host port to ingest syslog messages. The port is not set by default. When required, the recommended value is 5144 or based on the port availability on the host | Optional |
netVizPort |
When > 0, the network visibility agent will be deployed in a sidecar along with the machine agent. By default the network visibility agent works with port 3892 |
Not set by default |
netVizImage |
Reference of the Network Agent image | "appdynamics/machine-agent-netviz:latest" |
proxyUrl |
Url of the proxy server (protocol://domain:port") | Optional |
proxyUser |
Proxy user credentials (user@password) | Optional |
propertyBag |
A string with any other machine agent parameters | Optional |
pks |
PKS deployment flag. Must be used on PKS environments | Optional |
enableMasters |
When set to true server visibility will be provided for Master nodes. By default only Worker nodes are monitored. On managed Kubernetes providers the flag has no effect, as the Master plane is not accessible | Optional |
image |
The Machine Agent image | "appdynamics/machine-agent-analytics:latest" |
imageWin |
The Machine Agent image for Windows nodes | Required when nodeOS is set to all or windows |
imagePullSecret |
Name of the image pull secret | Optional |
nodeSelector |
Labels that identify nodes for scheduling of the daemonset pods | Optional |
tolerations |
A list of tolerations | Optional |
env |
List of environment variables | Optional |
args |
List of command arguments | Optional |
ports |
List of ports. All declared ports will be added to the service that fronts the Machine Agent pods | Optional |
resources |
Definitions of resources and limits for the machine agent | See example below |
priorityClassName |
Name of the priority class, e.g. system-node-critical |
Optional. If set, the infraviz resource and all dependencies including RBAC must be deployed to kube-system namespace |
Example resource limits:
resources:
limits:
cpu: 600m
memory: "1G"
requests:
cpu: 300m
memory: "800M"
- Server and Docker visibility
apiVersion: appdynamics.com/v1alpha1
kind: InfraViz
metadata:
name: appd-infraviz
namespace: appdynamics
spec:
controllerUrl: http://saas.appdynamics.com
image: docker.io/appdynamics/machine-agent-analytics:latest //default
account: customer1
globalAccount: customer1_f1d654a0-5
enableDockerViz: "true"
enableMasters: true // will deploy to master nodes too, if possible
stdoutLogging: true // log to console
nodeSelector:
kubernetes.io/os: linux
resources:
limits:
cpu: 500m
memory: "1G"
requests:
cpu: 200m
memory: "800M"
- Server and Network visibility
apiVersion: appdynamics.com/v1alpha1
kind: InfraViz
metadata:
name: appd-infraviz
namespace: appdynamics
spec:
controllerUrl: http://saas.appdynamics.com
image: appdynamics/machine-agent-analytics:latest
account: customer1
globalAccount: customer1_f1d654a0-5
enableDockerViz: "true"
enableMasters: true
stdoutLogging: true
netVizImage: appdynamics/machine-agent-netviz:latest // by default
netVizPort: 3892 //setting the port enables NetViz
nodeSelector:
kubernetes.io/os: linux
resources:
limits:
cpu: 500m
memory: "1G"
requests:
cpu: 200m
memory: "800M"
- Customizing ports. Syslog, Analytics
apiVersion: appdynamics.com/v1alpha1
kind: InfraViz
metadata:
name: appd-infraviz
namespace: appdynamics
spec:
controllerUrl: http://saas.appdynamics.com
image: appdynamics/machine-agent-analytics:latest
account: customer1
globalAccount: customer1_f1d654a0-5
enableDockerViz: "true"
enableMasters: true
stdoutLogging: true
syslogPort: 5144
biqPort: 9090 // by default
nodeSelector:
kubernetes.io/os: linux
resources:
limits:
cpu: 500m
memory: "1G"
requests:
cpu: 200m
memory: "800M"
- Deploy to Windows nodes
apiVersion: appdynamics.com/v1alpha1
kind: InfraViz
metadata:
name: appd-infraviz
namespace: appdynamics
spec:
controllerUrl: http://saas.appdynamics.com
account: customer1
globalAccount: customer1_f1d654a0-5
imageWin: docker.io/appdynamics/machine-agent-analytics:20.6.0-win-ltsc2019
nodeOS: windows
enableMasters: false
stdoutLogging: true
resources:
limits:
cpu: 500m
memory: "1G"
requests:
cpu: 200m
memory: "800M"
- Deploy to Mixed OS Clusters
apiVersion: appdynamics.com/v1alpha1
kind: InfraViz
metadata:
name: appd-infraviz
namespace: appdynamics
spec:
controllerUrl: http://saas.appdynamics.com
account: customer1
globalAccount: customer1_f1d654a0-5
image: "docker.io/appdynamics/machine-agent-analytics:latest"
imageWin: docker.io/appdynamics/machine-agent-analytics:20.6.0-win-ltsc2019
nodeOS: all
enableMasters: true
stdoutLogging: true
resources:
limits:
cpu: 500m
memory: "1G"
requests:
cpu: 200m
memory: "800M"
- Deploy to Infra nodes
apiVersion: appdynamics.com/v1alpha1
kind: InfraViz
metadata:
name: appd-infraviz
namespace: appdynamics
spec:
controllerUrl: http://saas.appdynamics.com
image: docker.io/appdynamics/machine-agent-analytics:latest
account: customer1
globalAccount: customer1_f1d654a0-5
enableDockerViz: "true"
enableMasters: true
stdoutLogging: true
nodeSelector:
kubernetes.io/os: linux
tolerations:
- effect: NoSchedule
key: node-role.kubernetes.io/infra
operator: Exists
resources:
limits:
cpu: 500m
memory: "1G"
requests:
cpu: 200m
memory: "800M"