How To: Connecting your Clusters / Installing the Komodor Agent

Komodor agents are deployed as a daemonset in every cluster with a simple Helm command, allowing for the monitoring of every Pod, Node, Workload, Job and Deployment.

Cloud and Distribution Requirements

Agents can be installed in any datacenter or cloud provider, any distribution. Komodor is provider and Distribution agnostic.
- EKS, AKS, GKE, OpenShift, Rancher, roll-your-own in a VM, on-prem
Komodor agents are required to have network access to communicate over an encrypted TLS connection to our SaaS platform

Can Komodor be installed in a fully air-gapped environment?

Cluster agents are required to communicate with the Komodor application layer hosted as SaaS in AWS.
If your Clusters are air-gapped in the strictest sense of the definition - zero access to the outside internet, ever - then Komodor would not work correctly.

Prerequisites for installation

1) Mirroring Images

For best performance, we recommend you mirror all images hosted under public.ecr.aws/komodor-public to your local image repository
If you are unable to access ECR, we also host in Dockerhub

2) Local Helm

Komodor agents are installed via Helm chart

helm repo add komodorio https://helm-charts.komodor.io ; helm repo update

If you are mirroring charts, your friendly local helm chart will need an update for imageRepo in this location.

3) Network / firewall requirements

Ensure firewall rules allow images to be pulled from public.ecr.aws/komodor-public or docker hub
Ensure app.komodor.com and telemetry.komodor.com are allowlisted.
Ensure websocket connections are allowed between komodor-agent and app.komodor.com (port 443)
Ensure any image scanning rules whitelist komodor-agent

4) Run the Memory Planning Utility

Before installing the Komodor agent, we recommend running our memory checker utility to analyze your cluster's resource requirements and determine appropriate memory limits. This helps prevent out-of-memory issues and ensures optimal performance.

Running the Utility - Quick Memory Analysis

# Clone or download and apply the memory planning utility resources - They will be installed in the 'komodor-precheck' namespace
kubectl apply -f https://raw.githubusercontent.com/komodorio/helm-charts/master/charts/komodor-agent/utilities/memory-planning/01-namespace.yaml
kubectl apply -f https://raw.githubusercontent.com/komodorio/helm-charts/master/charts/komodor-agent/utilities/memory-planning/02-configmap.yaml
kubectl apply -f https://raw.githubusercontent.com/komodorio/helm-charts/master/charts/komodor-agent/utilities/memory-planning/03-serviceaccount.yaml
kubectl apply -f https://raw.githubusercontent.com/komodorio/helm-charts/master/charts/komodor-agent/utilities/memory-planning/04-clusterrole.yaml
kubectl apply -f https://raw.githubusercontent.com/komodorio/helm-charts/master/charts/komodor-agent/utilities/memory-planning/05-clusterrolebinding.yaml
kubectl apply -f https://raw.githubusercontent.com/komodorio/helm-charts/master/charts/komodor-agent/utilities/memory-planning/06-job.yaml

# Monitor the analysis
kubectl logs -f job/komodor-memory-checker -n komodor-precheck

# View results and recommendations
kubectl logs job/komodor-memory-checker -n komodor-precheck | grep -A 10 "MEMORY RECOMMENDATIONS"

# Clean up after analysis
kubectl delete namespace komodor-precheck

Using the Results

Instructions for how to apply these results to your agent configuration are in the section below
For more configuration options, see the Memory Planning Utility Documentation.

Installing the Komodor Agent

To connect a cluster, the user either needs the OOTB Account-Admin role or a custom policy that includes 'manage:agents'.

Adding Clusters via the UI

If there are Zero clusters in your account, there will be a large prompt to add a cluster when you first login into the UI

Select ‘Connect new cluster’ under the Clusters scope from the left navigation panel
In the 'Set Up Komodor' popup, populate a cluster name. What you add here is what the cluster's display name will be for all users in Komodor.
Clicking 'Next' results in a Helm command in the format below. Open up a terminal session to your cluster and run

helm install komodor-agent komodorio/komodor-agent --set apiKey=akeyhere --set clusterName=whatever-you-like

Note: The API key that is generated in from the UI in this command is unique to your account and can be used programmatically

To apply the memory recommendations from the Memory Planning utility in your Helm install command:

helm upgrade --install komodor-agent komodorio/komodor-agent \
  --set apiKey=<YOUR_API_KEY_HERE> \
  --set clusterName=<CLUSTER_NAME> \
  --set components.komodorAgent.watcher.resources.requests.memory=<RECOMMENDED_REQUEST> \
  --set components.komodorAgent.watcher.resources.limits.memory=<RECOMMENDED_LIMIT>

Adding Clusters programmatically

After connecting the 1st account via UI, the API key returned can be used to install the agent in multiple clusters via CI/CD or bash automation with the Helm command

For ArgoCD, here is a template you may use

apiVersion: v1
kind: Secret
metadata:
  name: komodor-agent-secrets
  namespace: komodor
type: Opaque
stringData:
  apiKey: POPULATE_API_KEY_HERE # Get API Key from https://app.komodor.com/settings/agents

---

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: komodor-agent
  namespace: argocd
spec:
  project: default
  source:
    repoURL: https://helm-charts.komodor.io/
    targetRevision: 2.12.0
    chart: komodor-agent
    helm:
      values: |
        clusterName: "POPULATE_CLUSTER_NAME_HERE"
        apiKeySecret: komodor-agent-secrets
  destination:
    server: https://kubernetes.default.svc
    namespace: komodor
  syncPolicy:
    automated:
      prune: true
      selfHeal: true
    syncOptions:
      - CreateNamespace=true
Collapse

Advanced Agent Configuration

The Advanced Configuration Guide
- Data Redaction

Deleting a Komodor Agent from a Cluster

Uninstall the deployment via the Helm chart

helm uninstall komodor-agent