Deploy SonarQube on Kubernetes
Kubernetes Deployment is only available for Community, Developer, and Enterprise Editions.
Deploying SonarQube on Kubernetes is still in the early phases. We've only tested deployment with the following recommendations and constraints, and deployment has some limitations as documented below.
You can find the SonarQube Helm chart on GitHub.
Your feedback is welcome at our community forum.
Kubernetes Environment Recommendations
When you want to operate SonarQube on Kubernetes, consider the following recommendations.
|Kubernetes Version||Helm Chart Version||SonarQube Version|
Pod Security Policies
The following widely-used Pod Security Policies cannot be used in combination with SonarQube:
- Privileged - The SonarQube images are currently intended to start as root in order to provision the PVC and drop to lower privileges after that.
- ReadOnlyFileSystem - SonarQube is doing some filesystem operations to the container filesystem in order to deploy the correct language analyzers and community plugins.
- MustRunAsNonRoot - There is a init container that needs to run privileged to ensure that the Elasticsearch requirements to the specific node are fulfilled.
Taints and Tolerations
We recommend binding SonarQube to a specific node and reserving this node for SonarQube. It greatly increases the stability of the service.
The following sections detail creating a taint on a specific node and letting the SonarQube deployment ignore this taint using a flag in the
values.yaml of the Helm Chart.
Creating a taint
In order to create a taint, you need to select a node that you want to reserve for SonarQube. Use the following command to get a list of all nodes attached to your Kubernetes Cluster:
kubectl get nodes
Select a node from the output of this command, and create a custom taint using the following command:
kubectl taint nodes <node> sonarqube=true:NoSchedule
This taint ensures that no additional pods are scheduled on this node.
Ignoring this Taint for SonarQube
To let the SonarQube deployment ignore the previously created taint, add the following section to the
tolerations: - key: "sonarqube" operator: "Equal" value: "true" effect: "NoSchedule"
Depending on your taint's name, you may need to adjust the key accordingly.
As described in the Taints and Tolerations section above, for stability, we recommend binding the SonarQube deployment to one node in your cluster. With one node now reserved for SonarQube, you need to label this node to be selected by the Kube-scheduler in the pod assignment.
Label a Node
Label the node for which you previously defined a taint with the following command:
kubectl label node <node> sonarqube=true
Bind Deployment to Label
To only let SonarQube be scheduled on nodes with this specific label, add the following section to the
nodeSelector: sonarqube: "true"
By combining node selection with taints and tolerations, SonarQube can run alone on one specific node independently from the rest of your software in your Kubernetes cluster. This results in better stability and performance of SonarQube. For more information, see the official Kubernetes documentation.
Node affinity and anti-affinity can be used in a way similar to node selectors but with more operators to choose from. However, we generally don’t recommend using this in combination with SonarQube as it can lead to recurring rescheduling of the SonarQube pod. If you still want to use affinity and anti-affinity, see the official Kubernetes documentation.
Helm chart specifics
We try to provide a good default with the Helm chart, but there are some points to consider while working with SonarQube on Kubernetes. Please read the following sections carefully to make the correct decisions for your environment.
Currently only helm3 is supported.
To install the Helm Chart from the GitHub Repository, you can use the following commands:
git clone https://github.com/SonarSource/helm-chart-sonarqube.git cd helm-chart-sonarqube helm dependency update kubectl create namespace sonarqube helm upgrade --install -f values.yaml -n sonarqube sonarqube ./
SonarQube comes with a bundled Elasticsearch and, as Elasticsearch is stateful, so is SonarQube. There is an option to persist the Elasticsearch indexes in a Persistent Volume, but with regular killing operations by the Kubernetes Cluster, these indexes can be corrupted. By default, persistency is disabled in the Helm chart.
Enabling persistency decreases the startup time of the SonarQube Pod significantly, but you are risking corrupting your Elasticsearch index. You can enable persistency by adding the following to the
persistence: enabled: true
Leaving persistency disabled results in a longer startup time until SonarQube is fully available, but you won't lose any data as SonarQube will persist all data in the database.
When you're working with your own CA or in an environment that uses self-signed certificates for your code repository platform, you can create a secret containing this certificate and add this certificate to the java truststore inside the SonarQube deployment directly during the deployment.
To enable this behavior, add the following to your
caCerts: secret: <secret name>
Get Certificate via openssl
If you already have a running installation of your code repository platform, you can extract the certificate with the following snippet using
echo -n | openssl s_client -connect <server url>:443 | sed -ne '/-BEGIN CERTIFICATE-/,/-END CERTIFICATE-/p' > cert.pem
This certificate needs to be Base64 encoded in order to be added as secret data.
Create base64 string cat cert.pem | base64 | tr -d "\n"
Note that you can also use
string-data here if you don't want to encode your certificate.
The Base64 encoded certificate can be added to the secret's data:
apiVersion: v1 kind: Secret metadata: name: <secret name> namespace: <sonarqube namespace> data: cert: <base64 string>
Then, create the secret in your Kubernetes cluster with the following command:
kubectl apply -f secret.yaml
To make the SonarQube service accessible from outside of your cluster, you most likely need an ingress. Creating a new ingress is also covered by the Helm chart. See the following section for help with creating one.
The SonarSource Helm chart has an optional dependency to the NGINX-ingress helm chart. If you already have NGINX-ingress present in your cluster, you can use it.
If you want to install NGINX as well, add the following to your
nginx: enabled: true
We recommend using the
ingress-class NGINX with a body size of at least 8MB. This can be achieved with the following changes to your
ingress: enabled: true # Used to create an Ingress record. hosts: - name: <Your Sonarqube FQDN> # Different clouds or configurations might need /* as the default path path: / # For additional control over serviceName and servicePort # serviceName: someService # servicePort: somePort annotations: kubernetes.io/ingress.class: nginx nginx.ingress.kubernetes.io/proxy-body-size: "8m"
Other Configuration Options
While we only document the most pressing Helm chart customizations in this documentation, there are other possibilities for you to choose to Customize the Chart Before Installing. Please see the Helm chart README file for more information on these.
As SonarQube is intended to be run anywhere, there are some drawbacks that are currently known when operating in Kubernetes. This list is not comprehensive, but something to keep in mind and points for us to improve on.
No Sidecar Support
There is currently no support for additional sidecar containers and, as a result, there is no support for log collection. SonarQube will print the main application log to stdout, but logs on the web, ce, or search component will be printed to separate file streams inside the container. If you want to use a sidecar container with the SonarQube deployment, you have to manually alter the deployment.
No Log Complete Collection
As previously mentioned, there's currently no support for a log collection to make SonarQube observable. Logs are printed to separate file streams as plaintext. If you still want to scrape these logs, you will need to manually alter the deployment to read these 4 file streams and send them to your log collection solution manually.
Readiness and Startup delays
When persistence is disabled, SonarQube startup takes significantly longer as the Elasticsearch indexes need to be rebuilt. As this delay depends on the amount of data in your SonarQube instance, the values for the startup/readiness and liveness probes need to be adjusted to your environment. We also recommend taking a look at the default limits for the SonarQube deployment as the amount of CPU available to SonarQube also impacts the startup time.
Problems with Azure Fileshare PVC
Currently, there is a known limitation when working on AKS that resonates around the use of Azure Fileshare. We recommend using another storage class for persistency on AKS.
Currently, no cloud-native monitoring solutions play nicely with SonarQube or are supported by SonarSource. It is, however, possible to expose at least the JMX metrics to Prometheus with the help of the Prometheus JMX exporter.
To use this option, set the following values in your
prometheusExporter: enabled: true config: rules: - pattern: ".*"
This downloads the Prometheus JMX exporter agent and adds it to the startup options of SonarQube. With this default configuration, the JMX metrics will be exposed on /metrics for Prometheus to scrape.
The config scope here defines a configuration that is understandable by the Prometheus JMX exporter. For more information, please see the documentation.