Running SonarQube as a Cluster is only possible with a Data Center Edition .
The Data Center Edition allows SonarQube to run in a clustered configuration to make it resilient to failures. For more details about the Data Center (DC) Edition, please see this page.
You need five servers dedicated to SonarQube. Servers can be VMs, it's not necessary to have physical machines.
Servers must be co-located (geographical redundancy is not supported).
You can find the operating system requirements for servers in the Requirements page.
The Data Center Edition supports PostgreSQL, Oracle, and Microsoft SQL Server. If your data is currently stored in MySQL you can use the Sonar DB Copy Tool to move it.
In addition to the five SonarQube servers, you must configure a reverse proxy / load balancer to load balance traffic between the two application nodes. The precise configuration of the reverse proxy / load balancer will vary by vendor, but the SonarQube DC Edition requirements are very simple:
Share requests (load) between the two application nodes configured in the SonarQube cluster
If you are using HTTPS, ensure you are meeting the requirements set out in Securing SonarQube Behind a Proxy
Note: there is no requirement for the load balancer to preserve sessions; this is handled by the in-built JWT mechanism
You need a dedicated license to activate the DC Edition. If you don't have it yet, please contact the SonarSource Sales Team.
Don't start this journey alone; as a Data Center Edition customer SonarSource will assist with the setup and configuration of your cluster. Get in touch with SonarSource Support for help.
There are two types of nodes:
an application node responsible for handling web requests from users (WebServer process) and handling analysis reports (ComputeEngine process)
a search node that is an Elasticsearch process that will store indices of data
In order to achieve high availability, the only supported configuration for the DC Edition comprises 5 servers:
2 application nodes containing both WebServer and ComputeEngine
3 search nodes that host Elasticsearch. For performance reasons, SSD are significantly better than HDD for these nodes
With this configuration, a node can be lost without impacting the service. More precisely, one application node and one search node can be lost without impacting users.
Here is a schema for the supported topology:
Here are the type of machines we used to perform our validation with a 200M Issues database. This could be used as a minimum recommendation to build your cluster.
Prepare your personalized SonarQube package:
Install the SonarQube Data Center Edition, following the usual Installing the Server documentation
Install all the other plugins you may want, confirm the connectivity with the DB is working well. If you do this step manually be sure to check compatibility with your SonarQube version using the Plugin Version Matrix
Zip the directory $SONARQUBE_HOME
You now have your own personalized SonarQube Data Center Edition package
Deploy your SonarQube package on the 4 other nodes:
It is expected that all application nodes are identical in term of hardware and software (same JVM build). Similarly, all search nodes should be identical to each other. But application and search nodes can differ. Generally, search nodes are configured with more CPU and RAM than application nodes.
Unzip your personalized SonarQube package from the previous step on the 4 others nodes
Congratulations, you now have 2 applications nodes and 3 search nodes completely identical in term of SonarQube softwares
Now you have 5 machines with the same SonarQube software it's time to start configuring them to specialize them. Some will become application nodes, some search nodes.
The idea is as follows: you need to edit the sonar.properties file on each node to configure the node's specialization. In the following example we will assume:
The VMs (server1, server2) having IP addresses ip1 and ip2 are going to be application nodes
The VMs having IP addresses ip3, ip4 and ip5 (server3, server4 and server5) are going to be search nodes
The default configuration to be applied for each node is the following:
The full set of cluster parameters is listed here.
Once this configuration is done, take a break and a coffee, then you can Operate your Cluster.