Building software is the primary use case for Jenkins. To accomplish that, Jenkins invokes build scripts containing user-specified code, and ensuring this is done in a safe manner must be a major concern for Jenkins administrators. The following sections discuss potential issues and strategies to deal with them.
As discussed in Controller Isolation, Jenkins should be set up for distributed builds to ensure build scripts are not executing on the controller, where they could access the Jenkins home directory or otherwise interfere with the operation of Jenkins.
An additional concern is that builds need to be isolated from each other: While simple Jenkins instances building a single project may be able to operate securely with a single static agent, administrators of more complex instances need to consider whether builds need to be isolated from each other. Some examples:
Jenkins hosts multiple teams, which are not allowed to access each others' projects or source code.
Jenkins builds pull requests/merge requests submitted to public repositories, e.g., on GitHub.
In these scenarios, build scripts running on the same agents may be able to access data related to other, unrelated, projects, and potentially even manipulate their build results. Even not assuming malicious intent, badly set up build scripts can accidentally delete unrelated data or cause resource exhaustion on the static agent, delaying builds.
Solutions to this issue include the following:
Use cloud providers that create a new agent for every build.
Plugins such as Job Restrictions Plugin limit which jobs can be run on certain nodes.
When using organization folders or multibranch Pipelines, Jenkins automatically builds new pull requests by default. Especially when a Jenkins instance builds projects from public repositories, this can open the door for abuse, such as:
Spam pull requests that trigger many new builds.
Malicious pull requests that modify build scripts, e.g., to mine cryptocurrency.
Strategies for dealing with this include:
Use cloud providers and limit how many agents can be launched in parallel. This is easy to implement, but can delay legitimate builds due to resource constraints.
Define timeouts in the Pipeline definition (typically
Jenkinsfile), as it cannot be modified by users without commit access given appropriate Pipeline trust settings.
Do not automatically build new pull requests and instead require committers to opt into them, e.g., using plugins like GitHub Label Filter
Build scripts and tools invoked during the build may change behavior depending on environment variables defined for the build. These variables can come from various sources, such as build parameters, predefined agent environment variables, files read from the workspace and injected into the environment, etc.
Handling Environment Variables discusses this topic in detail, including solutions.
Please submit your feedback about this page through this quick form.
Alternatively, if you don't wish to complete the quick form, you can simply indicate if you found this page helpful?
See existing feedback here.