1
0
mirror of https://github.com/open-telemetry/opentelemetry-go.git synced 2024-12-12 10:04:29 +02:00
opentelemetry-go/example/otel-collector
Hui Kang d75ccbb378
Add a Shutdown method to sdk TraceProvider (#1227)
* Add a Shutdown method to api TraceProvider

- sdktraceprovider shutdown span processors
- In examples, replace processosr shutdown with
  traceprovider's shutdown

Signed-off-by: Hui Kang <kangh@us.ibm.com>

* remove shutdown in the api provider interface

* Add context in parameter and return error

* handle error in shutdown

* Update CHANGELOG.md

Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
2020-10-26 09:20:49 -07:00
..
k8s fix otel collector example (#1006) 2020-07-31 11:34:46 -07:00
go.mod Pre release v0.13.0 (#1231) 2020-10-08 18:18:50 -04:00
go.sum Bump google.golang.org/grpc from 1.31.1 to 1.32.0 in /sdk (#1145) 2020-09-09 13:25:27 -04:00
main.go Add a Shutdown method to sdk TraceProvider (#1227) 2020-10-26 09:20:49 -07:00
Makefile Merge otlp collector examples (#841) 2020-06-23 08:37:07 -07:00
README.md Merge otlp collector examples (#841) 2020-06-23 08:37:07 -07:00

OpenTelemetry Collector Traces Example

This example illustrates how to export trace and metric data from the OpenTelemetry-Go SDK to the OpenTelemetry Collector. From there, we bring the trace data to Jaeger and the metric data to Prometheus The complete flow is:

                                          -----> Jaeger (trace)
App + SDK ---> OpenTelemtry Collector ---|
                                          -----> Prometheus (metrics)

Prerequisites

You will need access to a Kubernetes cluster for this demo. We use a local instance of microk8s, but please feel free to pick your favorite. If you do decide to use microk8s, please ensure that dns and storage addons are enabled

microk8s enable dns storage

For simplicity, the demo application is not part of the k8s cluster, and will access the OpenTelemetry Collector through a NodePort on the cluster. Note that the NodePort opened by this demo is not secured.

Ideally you'd want to either have your application running as part of the kubernetes cluster, or use a secured connection (NodePort/LoadBalancer with TLS or an ingress extension).

Deploying to Kubernetes

All the necessary Kubernetes deployment files are available in this demo, in the k8s folder. For your convenience, we assembled a makefile with deployment commands (see below). For those with subtly different systems, you are, of course, welcome to poke inside the Makefile and run the commands manually. If you use microk8s and alias microk8s kubectl to kubectl, the Makefile will not recognize the alias, and so the commands will have to be run manually.

Setting up the Prometheus operator

If you're using microk8s like us, simply do

microk8s enable prometheus

and you're good to go. Move on to Using the makefile.

Otherwise, obtain a copy of the Prometheus Operator stack from coreos:

git clone https://github.com/coreos/kube-prometheus.git
cd kube-prometheus
kubectl create -f manifests/setup

# wait for namespaces and CRDs to become available, then
kubectl create -f manifests/

And to tear down the stack when you're finished:

kubectl delete --ignore-not-found=true -f manifests/ -f manifests/setup

Using the makefile

Next, we can deploy our Jaeger instance, Prometheus monitor, and Collector using the makefile.

# Create the namespace
make namespace-k8s

# Deploy Jaeger operator
make jaeger-operator-k8s

# After the operator is deployed, create the Jaeger instance
make jaeger-k8s

# Then the Prometheus instance. Ensure you have enabled a Prometheus operator
# before executing (see above).
make prometheus-k8s

# Finally, deploy the OpenTelemetry Collector
make otel-collector-k8s

If you want to clean up after this, you can use the make clean-k8s to delete all the resources created above. Note that this will not remove the namespace. Because Kubernetes sometimes gets stuck when removing namespaces, please remove this namespace manually after all the resources inside have been deleted, for example with

kubectl delete namespaces observability

Configuring the OpenTelemetry Collector

Although the above steps should deploy and configure everything, let's spend some time on the configuration of the Collector.

One important part here is that, in order to enable our application to send data to the OpenTelemetry Collector, we need to first configure the otlp receiver:

...
  otel-collector-config: |
    receivers:
      # Make sure to add the otlp receiver.
      # This will open up the receiver on port 55680.
      otlp:
        endpoint: 0.0.0.0:55680
    processors:    
...

This will create the receiver on the Collector side, and open up port 55680 for receiving traces.

The rest of the configuration is quite standard, with the only mention that we need to create the Jaeger and Prometheus exporters:

...
    exporters:
      jaeger_grpc:
        endpoint: "jaeger-collector.observability.svc.cluster.local:14250"

      prometheus:
           endpoint: 0.0.0.0:8889
           namespace: "testapp"
...

OpenTelemetry Collector service

One more aspect in the OpenTelemetry Collector configuration worth looking at is the NodePort service used for accessing it:

apiVersion: v1
kind: Service
metadata:
        ...
spec:
  ports:
  - name: otlp # Default endpoint for otlp receiver.
    port: 55680
    protocol: TCP
    targetPort: 55680
    nodePort: 30080
  - name: metrics # Endpoint for metrics from our app.
    port: 8889
    protocol: TCP
    targetPort: 8889
  selector:
    component: otel-collector
  type:
    NodePort

This service will bind the 55680 port used to access the otlp receiver to port 30080 on your cluster's node. By doing so, it makes it possible for us to access the Collector by using the static address <node-ip>:30080. In case you are running a local cluster, this will be localhost:30080. Note that you can also change this to a LoadBalancer or have an ingress extension for accessing the service.

Running the code

You can find the complete code for this example in the main.go file. To run it, ensure you have a somewhat recent version of Go (preferably >= 1.13) and do

go run main.go

The example simulates an application, hard at work, computing for ten seconds then finishing.

Viewing instrumentation data

Now the exciting part! Let's check out the telemetry data generated by our sample application

Jaeger UI

First, we need to enable an ingress provider. If you've been using microk8s, do

microk8s enable ingress

Then find out where the Jaeger console is living:

kubectl get ingress --all-namespaces

For us, we get the output

NAMESPACE       NAME           CLASS    HOSTS   ADDRESS     PORTS   AGE
observability   jaeger-query   <none>   *       127.0.0.1   80      5h40m

indicating that the Jaeger UI is available at http://localhost:80. Navigate there in your favorite web-browser to view the generated traces.

Prometheus

Unfortunately, the Prometheus operator doesn't provide a convenient out-of-the-box ingress route for us to use, so we'll use port-forwarding instead. Note: this is a quick-and-dirty solution for the sake of example. You will be attacked by shady people if you do this in production!

kubectl --namespace monitoring port-forward svc/prometheus-k8s 9090

Then navigate to http://localhost:9090 to view the Prometheus dashboard.