What happened:

The subscriber-config ConfigMap is initialized with a hardcoded localhost:32506 URL during Litmus installation, causing the subscriber pod to crash with CrashLoopBackOff. The pod cannot connect to the Litmus server because localhost doesn't resolve inside Kubernetes pods.

Error observed:

time="2026-01-03T04:16:11Z" level=fatal msg="Failed to confirm agent" 
error="Post \"http://localhost:32506/api/query\": dial tcp [::1]:32506: connect: connection refused"

Pod status:

NAME                          READY   STATUS             RESTARTS   AGE
subscriber-77764cd5cb-xxxxx   0/1     CrashLoopBackOff   5          5m

What you expected to happen:

The subscriber-config ConfigMap should be initialized with the correct Kubernetes service name and port:

• Service Name: chaos-litmus-server-service (or appropriate service name)
• Port: 9002 (GraphQL server port from the service)
• Endpoint: /query (GraphQL query endpoint)

Expected ConfigMap value:

SERVER_ADDR: http://chaos-litmus-server-service:9002/query

Where can this issue be corrected? (optional)

The issue appears to be in the Litmus installation/initialization code that creates the subscriber-config ConfigMap. This could be in:

• Helm chart templates (if using Helm installation)
• Operator initialization code
• ConfigMap creation logic in the installation manifests

The ConfigMap is created in the litmus namespace with the name subscriber-config.

How to reproduce it (as minimally and precisely as possible):

1. Install Litmus 3.24.0 using Helm or kubectl apply:

   # Using Helm
   helm repo add litmuschaos https://litmuschaos.github.io/litmus-helm/
   helm install chaos litmuschaos/litmus --namespace=litmus --create-namespace
   
   # OR using kubectl (apply the provided YAML)
   kubectl apply -f LitmusChaos-Local-Development.yml

2. Check the subscriber-config ConfigMap:

   kubectl get configmap subscriber-config -n litmus -o yaml

3. Observe the incorrect SERVER_ADDR value:

   data:
     SERVER_ADDR: http://localhost:32506/api/query

4. Check subscriber pod status:

   kubectl get pods -n litmus -l app=subscriber

5. Check subscriber pod logs:

   kubectl logs -n litmus -l app=subscriber --tail=20

6. Expected result: Pod will be in CrashLoopBackOff state with connection refused errors.

Environment Details:

• Litmus Version: 3.24.0
• Kubernetes Version: Rancher Desktop (K3s)
• Platform: macOS (ARM64/Apple Silicon)
• Installation Method: kubectl apply (using provided YAML file)
• Namespace: litmus

Workaround:

Manually patch the ConfigMap to use the correct service name:

# Get the correct service name and port
kubectl get svc -n litmus | grep litmus-server

# Patch the ConfigMap
kubectl patch configmap subscriber-config -n litmus --type merge \
  -p '{"data":{"SERVER_ADDR":"http://chaos-litmus-server-service:9002/query"}}'

# Delete the subscriber pod to restart with new config
kubectl delete pod -n litmus -l app=subscriber

# Verify the pod is now running
kubectl get pods -n litmus -l app=subscriber
kubectl logs -n litmus -l app=subscriber --tail=20

After the workaround, the subscriber should show:

AgentID: 8c97b69d-7bee-4802-8c14-fbf3b4bb1d1c has been confirmed
Connecting to ws://chaos-litmus-server-service:9002/query
Server connection established, Listening....

Anything else we need to know?:

1. Root Cause Analysis:

   • The ConfigMap annotation shows this value was set during installation: "SERVER_ADDR":"http://localhost:32506/api/query"
   • This appears to be a default value bug in the installation process
   • The value 32506 is a NodePort (external access), not the internal service port
   • The endpoint /api/query is incorrect (should be /query)

2. Impact:

   • Subscriber pod cannot connect to server
   • Infrastructure connection fails in Litmus UI
   • No communication between subscriber and server
   • Infrastructure status shows as disconnected
   • This affects all fresh installations of Litmus 3.24.0

3. Service Information:

   # Actual service configuration
   kubectl get svc chaos-litmus-server-service -n litmus
   # Shows ports: 9002 (GraphQL server) and 8000 (RPC server)

4. Related Components:

   • ConfigMap: subscriber-config in litmus namespace
   • Deployment: subscriber in litmus namespace
   • Service: chaos-litmus-server-service in litmus namespace

5. Additional Notes:

   • This bug affects the subscriber component specifically
   • Other components (operator, exporter, event-tracker) work correctly
   • The bug is reproducible on fresh installations
   • The workaround is effective but should not be necessary