Cluster-level Kubernetes Logging with Honeycomb
Honeycomb's Kubernetes agent aggregates logs across a Kubernetes cluster. Stop managing log storage in all your clusters and start tracking down real problems.
To get started with Honeycomb, check out the Honeycomb general quickstart.
This README includes some basic information about getting started with the Kubernetes agent. Please see Honeycomb's Kubernetes documentation for more comprehensive documentation.
How it Works
honeycomb-agent runs as a DaemonSet on each node in a cluster. It reads container log files from the node's filesystem, augments them with metadata from the Kubernetes API, and ships them to Honeycomb so that you can see what's going on.
The following steps will deploy the Honeycomb agent to each node in your cluster, and configure it to process logs from all pods.
Grab your Honeycomb writekey from your account page, and create a Kubernetes secret from it:
kubectl create secret generic honeycomb --from-literal=api-key=$WRITEKEY --namespace=honeycomb
Run the agent
kubectl apply -f examples/quickstart.yaml
This will do three things:
- create a service account for the agent so that it can list pods from the API
- create a minimal
ConfigMapcontaining configuration for the agent
- create a DaemonSet from the agent.
It's best if all of your containers output structured JSON logs. But that's not always realistic. In particular, you're likely to operate third-party services, such as proxies or databases, that don't log JSON.
You may also want to aggregate logs from specific services, rather than from everything that might be running in a cluster.
In order to get usefully structured data from services, you can use Kubernetes label selectors to describe how to parse logs for specific services.
For example, to parse logs from pods with the label
app: nginx as NGINX logs,
you'd specify the following configuration:
watchers: - labelSelector: "app=nginx" dataset: kubernetes-nginx parser: nginx
You might want to do additional munging of events before sending them to
Honeycomb. For each label selector, you can specify a list of
which will be applied in order. For example:
watchers: - labelSelector: "app=nginx" parser: nginx dataset: kubernetes-nginx processors: - request_shape: # Unpack the field "request": "GET /path HTTP/1.x" field: request # into its constituent components - drop_field: # Remove the "user_email" field from all events field: user_email - sample: # Sample events: only send one in 20 type: static rate: 20
See the docs for more examples.