Integration troubleshooting

Open in ChatGPTOpen in Claude

You installed the OpenTelemetry Integration chart, instrumented your applications, and nothing is arriving in Coralogix. Two things are usually responsible:

• The Helm installation failed, or the collectors are not healthy.
• Your application is not sending to the collector.

Work through the checks below in order. Each one narrows down which of the two it is.

Check the agent logs

1.

Find the agent pod

It appears once, with the installation name you chose:

kubectl get pods -o wide -n $NAMESPACE | grep opentelemetry-agent

Its STATUS should be Running. Pending means it has not scheduled yet. Rerun the command. AGE should match the time since your last Helm upgrade; anything much older suggests the upgrade did not take.

2.

Stream its logs

Copy the pod's full name from the output above. The --follow flag live tails the logs in your terminal:

kubectl logs --follow <opentelemetry-agent-pod-name> -n <namespace>

Check the cluster collector

Missing data, or an incomplete view such as absent Kubernetes events, usually points here.

1.

Confirm all three pods are running

The agent, the cluster collector, and Kube State Metrics all need to be up:

kubectl get pods -o wide -n $NAMESPACE
NAME                                                  READY   STATUS    RESTARTS   AGE   IP              NODE                                               NOMINATED NODE   READINESS GATES
coralogix-opentelemetry-agent-jhl6q                   1/1     Running   0          43m   172.31.27.209   ip-172-31-27-209.ap-southeast-1.compute.internal   <none>           <none>
coralogix-opentelemetry-collector-84d684d45c-2h4rt    1/1     Running   0          43m   172.31.31.122   ip-172-31-27-209.ap-southeast-1.compute.internal   <none>           <none>
otel-integration-kube-state-metrics-bbc57cd9b-9kgrp   1/1     Running   0          43m   172.31.18.170   ip-172-31-27-209.ap-southeast-1.compute.internal   <none>           <none>

If any is missing or not Running, pull its logs before going further.

2.

Confirm the extension package is installed

Check the Kubernetes OpenTelemetry extension in your Coralogix account is on the latest version, with all relevant Applications and Subsystems selected.

3.

Look for gaps in the dashboards

Select Grafana, then Dashboards, then Manage, and open the K8sOtel folder. A populated dashboard looks like the one below.

Shows a healthy dashboard. Panels reading No data instead mean either telemetry is not being exported from the collectors, or a component on the collector is misconfigured.

Test endpoint connectivity

Validate that the cluster can resolve your OpenTelemetry endpoint. This command creates a temporary busybox pod, runs nslookup against the endpoint, and cleans up after itself:

kubectl run busybox --image=busybox:1.28 --rm -it --restart=Never -- nslookup ingress.eu2.coralogix.com:443

A successful lookup returns the server and the resolved address. An error means the cluster cannot reach the domain server, a networking or DNS problem outside the Collector.

Check the chart configuration

Unexpected behavior, wrong telemetry, or a failed Helm install usually means values.yaml does not match the structure the chart expects, or your overrides did not apply.

1.

Compare against the defaults

Diff your file against the chart's default values.yaml.

2.

Inspect what actually deployed

The rendered ConfigMaps show the configuration as the collectors received it, overrides included:

kubectl get configmap -n $NAMESPACE
NAME                                DATA   AGE
coralogix-opentelemetry-agent       1      4d
coralogix-opentelemetry-collector   1      4d
kube-root-ca.crt                    1      118d

kubectl get configmap -n otel -o yaml coralogix-opentelemetry-agent
kubectl get configmap -n otel -o yaml coralogix-opentelemetry-collector

For background on how the overrides are templated, see Helm values files.

Next steps

Read the Kubernetes Observability using OpenTelemetry FAQs.