Navigation Menu

Skip to content
This repository has been archived by the owner on May 18, 2020. It is now read-only.


Repository files navigation

Jaeger OpenShift Templates

This repository is deprecated

The current recommended way of installing and managing Jaeger in a production OpenShift cluster is via the Jaeger Operator.

You can still use, report issues and send pull-requests against this repository, but not all features from the Operator are possible or will be backported to the templates from this repository here.

Use the Kubernetes templates if you need a quick start and don't want to install the Operator.

Development setup

This template uses an in-memory storage with a limited functionality for local testing and development. Do not use this template in production environments, although there are a number of parameters in the template to constrain the maximum number of traces and the amount of CPU/Memory consumed to prevent node instability.

Install everything in the current namespace:

oc process -f | oc create -f -

Once everything is ready, oc status tells you where to find Jaeger URL.

Production setup

Backing storage

The Jaeger Collector and Query require a backing storage to exist before being started up. As a starting point for your own templates, we provide a basic template to deploy Cassandra. It is not ready for production and should be adapted before any real usage.


To use our Cassandra template:

oc create -f production/cassandra.yml
oc create -f production/configmap-cassandra.yml

The Cassandra template includes also a Kubernetes Job that creates the schema required by the Jaeger components. It's advisable to wait for this job to finish before deploying the Jaeger components. To check the status of the job, run:

oc get job jaeger-cassandra-schema-job

The job should have 1 in the SUCCESSFUL column.


To use our Elasticsearch template:

oc create -f production/elasticsearch.yml
oc create -f production/configmap-elasticsearch.yml

The Elasticsearch template in this repository deploys only one node and overall it's not production quality! We encourage you to use other templates, for example docker-rhel-elasticsearch. This Elasticsearch deployment is also used by integration tests for this repository.

Jaeger configuration

The Jaeger Collector, Query and Agent require a ConfigMap to exist on the same namespace, named jaeger-configuration. This ConfigMap is included in the storage templates, as each backing storage have their own specific configuration entries, but in your environment, you'll probably manage it differently.

If changes are required for the configuration, the edit command can be used:

oc edit configmap jaeger-configuration

Jaeger components

The main production template deploys the Collector and the Query Service (with UI) as separate individually scalable services.

oc process -f production/jaeger-production-template.yml | oc create -f -

If the backing storage is not ready by the time the Collector/Agent start, they will fail and Kubernetes will reschedule the pod. It's advisable to either wait for the backing storage to estabilize, or to ignore such failures for the first few minutes.

Once everything is ready, the Jaeger Query service location can be discovered by running:

oc get route jaeger-query

It should be similar to:

Deploying the agent as sidecar

The Jaeger Agent is designed to be deployed local to your service, so that it can receive traces via UDP keeping your application's load minimal. As such, it's ideal to have the Agent to be deployed as a sidecar to your application's component, just add it as a container within any struct that supports spec.containers, like a Pod, Deployment and so on.

For instance, assuming that your application is named myapp and the image is for it is openshift/hello-openshift, your Deployment descriptor would be something like:

- apiVersion: extensions/v1beta1
  kind: Deployment
    name: myapp
          app: myapp
        - image: openshift/hello-openshift
          name: myapp
          - containerPort: 8080
        - image: jaegertracing/jaeger-agent
          name: jaeger-agent
          - containerPort: 5775
            protocol: UDP
          - containerPort: 5778
          - containerPort: 6831
            protocol: UDP
          - containerPort: 6832
            protocol: UDP
          - ""

The Jaeger Agent will then be available to your application at localhost:5775/localhost:6831/localhost:6832. In most cases, you don't need to specify a hostname or port to your Jaeger Tracer, as it will default to the right values already.

Persistent storage

Even though this template uses a stateful Cassandra, backing storage is set to emptyDir. It's more appropriate to create a PersistentVolumeClaim/PersistentVolume and use it instead. Note that this Cassandra deployment does not support deleting pods or scaling down, as this might require administrative tasks that are dependent on the final deployment architecture.

Exposing Jaeger-Collector outside of Cluster

Collector is by default accessible only to services running inside the cluster. The easiest approach to expose the collector outside of the cluster is via the jaeger-collector-http HTTP port using an OpenShift Route:

oc create route edge --service=jaeger-collector --port jaeger-collector-http --insecure-policy=Allow

This allows clients to send data directly to Collector via HTTP senders. If you want to use the Agent then use ExternalIP or NodePort to expose the Collector service.

Note that doing so will open the collector to be used by any external party, who will then be able to create arbitrary spans. It's advisable to put an OAuth Security Proxy in front of the collector and expose this proxy instead.

Using a different version

The templates are using a specific version and latest in all-in-one template. If you need to use a different Docker image version, specify it via the template parameter IMAGE_VERSION, as follows:

oc process -f <path-to-template> -p IMAGE_VERSION=<version> | oc create -n jaeger-infra -f -

A list of tags can be found here:

Getting an OpenShift cluster running

As a developer looking to try this out locally, the easiest is to use the oc cluster up command. Getting this command might be as easy as running dnf install origin-clients on a recent Fedora desktop. Refer to the OpenShift installation guide or quick start guide for more information. Another alternative is to use minishift.

OpenShift Service Catalog

The templates available in this repository are also suitable to be added to the OpenShift Service Catalog. To do so, run oc create -n openshift -f <template> and wait for the Template Service Broker to pick it up, which might take a couple of minutes.

oc create -n openshift -f
oc create -n openshift -f

Before provisioning a production instance, however, your users will need to provision a backing storage and install an appropriate ConfigMap, as described previously in this document.


If you need to remove the all Jaeger components created by this template, run:

oc delete all,template,configmap -l jaeger-infra


Tests are based on Arquillian Cube which require an active connection to openshift cluster (via oc). Currently all templates are tested on minishift or local all-in-one cluster (oc cluster up).

minishift start // or oc cluster up
mvn clean verify -Pe2e


Tracer may sometimes fail to resolve Jaeger's address. In such case run the following command and restart the affected applications.:

sudo iptables -F


Apache-2.0 License