• Workflow
• Connecting to the Kubernetes API
• The keywords
  • Configuring executor Service Account
  • Overwriting Kubernetes Namespace
  • Overwriting Kubernetes Default Service Account
  • Setting Bearer Token to be Used When Making Kubernetes API calls
  • Overwriting pod annotations
  • Overwriting Build Resources
• Define keywords in the configuration TOML
• Using volumes
  • Host Path volumes
  • PVC volumes
  • ConfigMap volumes
  • Secret volumes
  • Empty Dir volumes
• Using Security Context
• Using services
• Using Docker in your builds
  • Exposing /var/run/docker.sock
  • Using docker:dind
  • Not supplying Git
  • Resource separation
• Job execution
  • Using kaniko

The Kubernetes executor

GitLab Runner can use Kubernetes to run builds on a Kubernetes cluster. This is possible with the use of the Kubernetes executor.

The Kubernetes executor, when used with GitLab CI, connects to the Kubernetes API in the cluster creating a Pod for each GitLab CI Job. This Pod is made up of, at the very least, a build container, a helper container, and an additional container for each service defined in the .gitlab-ci.yml or config.toml files. The names for these containers are as follows:

• The build container is build
• The helper container is helper
• The services containers are svc-X where X is [0-9]+

Note that when services and containers are running in the same Kubernetes pod, they are all sharing the same localhost address. The following restrictions are then applicable:

• Since GitLab Runner 12.8 and Kubernetes 1.7, the services are accessible via their DNS names. If you are using an older version you will have to use localhost.
• You cannot use several services using the same port (e.g., you cannot have two mysql services at the same time).

Workflow

The Kubernetes executor divides the build into multiple steps:

1. Prepare: Create the Pod against the Kubernetes Cluster. This creates the containers required for the build and services to run.
2. Pre-build: Clone, restore cache and download artifacts from previous stages. This is run on a special container as part of the Pod.
3. Build: User build.
4. Post-build: Create cache, upload artifacts to GitLab. This also uses the special container as part of the Pod.

Connecting to the Kubernetes API

The following options are provided, which allow you to connect to the Kubernetes API:

• host: Optional Kubernetes apiserver host URL (auto-discovery attempted if not specified)
• cert_file: Optional Kubernetes apiserver user auth certificate
• key_file: Optional Kubernetes apiserver user auth private key
• ca_file: Optional Kubernetes apiserver ca certificate

The user account provided must have permission to create, list and attach to Pods in the specified namespace in order to function.

If you are running the GitLab CI Runner within the Kubernetes cluster you can omit all of the above fields to have the Runner auto-discover the Kubernetes API. This is the recommended approach.

If you are running it externally to the Cluster then you will need to set each of these keywords and make sure that the Runner has access to the Kubernetes API on the cluster.

The keywords

The following keywords help to define the behavior of the Runner within Kubernetes:

• namespace: Namespace in which to run Kubernetes Pods
• namespace_overwrite_allowed: Regular expression to validate the contents of the namespace overwrite environment variable (documented below). When empty, it disables the namespace overwrite feature
• privileged: Run containers with the privileged flag
• cpu_limit: The CPU allocation given to build containers
• cpu_limit_overwrite_max_allowed: The max amount the CPU allocation can be written to for build containers. When empty, it disables the cpu limit overwrite feature
• memory_limit: The amount of memory allocated to build containers
• memory_limit_overwrite_max_allowed: The max amount the memory allocation can be written to for build containers. When empty, it disables the memory limit overwrite feature
• service_cpu_limit: The CPU allocation given to build service containers
• service_memory_limit: The amount of memory allocated to build service containers
• helper_cpu_limit: The CPU allocation given to build helper containers
• helper_memory_limit: The amount of memory allocated to build helper containers
• cpu_request: The CPU allocation requested for build containers
• cpu_request_overwrite_max_allowed: The max amount the CPU allocation request can be written to for build containers. When empty, it disables the cpu request overwrite feature
• memory_request: The amount of memory requested from build containers
• memory_request_overwrite_max_allowed: The max amount the memory allocation request can be written to for build containers. When empty, it disables the memory request overwrite feature
• service_cpu_request: The CPU allocation requested for build service containers
• service_memory_request: The amount of memory requested for build service containers
• helper_cpu_request: The CPU allocation requested for build helper containers
• helper_memory_request: The amount of memory requested for build helper containers
• pull_policy: specify the image pull policy: never, if-not-present, always. The cluster’s image default pull policy will be used if not set.
  • See also if-not-present security considerations.
• node_selector: A table of key=value pairs of string=string. Setting this limits the creation of pods to Kubernetes nodes matching all the key=value pairs
• node_tolerations: A table of "key=value" = "Effect" pairs in the format of string=string:string. Setting this allows pods to schedule to nodes with all or a subset of tolerated taints. Only one toleration can be supplied through environment variable configuration. The key, value, and effect match with the corresponding field names in Kubernetes pod toleration configuration.
• image_pull_secrets: A array of secrets that are used to authenticate Docker image pulling
• helper_image: (Advanced) Override the default helper image used to clone repos and upload artifacts.
• terminationGracePeriodSeconds: Duration after the processes running in the pod are sent a termination signal and the time when the processes are forcibly halted with a kill signal
• poll_interval: How frequently, in seconds, the runner will poll the Kubernetes pod it has just created to check its status (default = 3).
• poll_timeout: The amount of time, in seconds, that needs to pass before the runner will time out attempting to connect to the container it has just created. Useful for queueing more builds that the cluster can handle at a time (default = 180).
• pod_labels: A set of labels to be added to each build pod created by the runner. The value of these can include environment variables for expansion.
• pod_annotations: A set of annotations to be added to each build pod created by the Runner. The value of these can include environment variables for expansion. Pod annotations can be overwritten in each build.
• pod_annotations_overwrite_allowed: Regular expression to validate the contents of the pod annotations overwrite environment variable. When empty, it disables the pod annotations overwrite feature
• pod_security_context: Configured through the configuration file, this sets a pod security context for the build pod. Read more about security context
• service_account: default service account to be used for making Kubernetes API calls.
• service_account_overwrite_allowed: Regular expression to validate the contents of the service account overwrite environment variable. When empty, it disables the service account overwrite feature
• bearer_token: Default bearer token used to launch build pods.
• bearer_token_overwrite_allowed: Boolean to allow projects to specify a bearer token that will be used to create the build pod.
• volumes: configured through the configuration file, the list of volumes that will be mounted in the build container. Read more about using volumes
• services: Since GitLab Runner 12.5, list of services attached to the build container using the sidecar pattern. Read more about using services.

Configuring executor Service Account

You can set the KUBERNETES_SERVICE_ACCOUNT environment variable or use --service-account flag.

Overwriting Kubernetes Namespace

Additionally, Kubernetes namespace can be overwritten on .gitlab-ci.yml file, by using the variable KUBERNETES_NAMESPACE_OVERWRITE.

This approach allows you to create a new isolated namespace dedicated for CI purposes, and deploy a custom set of Pods. The Pods spawned by the runner will take place on the overwritten namespace, for simple and straight forward access between containers during the CI stages.

variables:
  KUBERNETES_NAMESPACE_OVERWRITE: ci-${CI_COMMIT_REF_SLUG}

Furthermore, to ensure only designated namespaces will be used during CI runs, set the configuration namespace_overwrite_allowed with an appropriate regular expression. When left empty the overwrite behavior is disabled.

Overwriting Kubernetes Default Service Account

Additionally, the Kubernetes service account can be overwritten in the .gitlab-ci.yml file by using the variable KUBERNETES_SERVICE_ACCOUNT_OVERWRITE.

This approach allows you to specify a service account that is attached to the namespace, which is useful when dealing with complex RBAC configurations.

variables:
  KUBERNETES_SERVICE_ACCOUNT_OVERWRITE: ci-service-account

To ensure only designated service accounts will be used during CI runs, set the configuration service_account_overwrite_allowed or set the environment variable KUBERNETES_SERVICE_ACCOUNT_OVERWRITE_ALLOWED with an appropriate regular expression. When left empty the overwrite behavior is disabled.

Setting Bearer Token to be Used When Making Kubernetes API calls

In conjunction with setting the namespace and service account as mentioned above, you may set the bearer token used when making API calls to create the build pods. This will allow project owners to use project secret variables to specify a bearer token. When specifying the bearer token, you must set the Host configuration keyword.

variables:
  KUBERNETES_BEARER_TOKEN: thebearertokenfromanothernamespace

Overwriting pod annotations

Additionally, Kubernetes pod annotations can be overwritten on the .gitlab-ci.yml file, by using KUBERNETES_POD_ANNOTATIONS_* for variables and key=value for the value. The pod annotations will be overwritten to the key=value. Multiple annotations can be applied. For example:

variables:
  KUBERNETES_POD_ANNOTATIONS_1: "Key1=Val1"
  KUBERNETES_POD_ANNOTATIONS_2: "Key2=Val2"
  KUBERNETES_POD_ANNOTATIONS_3: "Key3=Val3"

Overwriting Build Resources

Additionally, Kubernetes CPU and memory allocations for requests and limits can be overwritten on the .gitlab-ci.yml file with the following variables:

 variables:
   KUBERNETES_CPU_REQUEST: 3
   KUBERNETES_CPU_LIMIT: 5
   KUBERNETES_MEMORY_REQUEST: 2Gi
   KUBERNETES_MEMORY_LIMIT: 4Gi

The values for these variables are restricted to what the max overwrite for that resource has been set to.

Define keywords in the configuration TOML

Each of the keywords can be defined in the config.toml for the GitLab Runner.

Here is an example config.toml:

concurrent = 4

[[runners]]
  name = "Kubernetes Runner"
  url = "https://gitlab.com/ci"
  token = "......"
  executor = "kubernetes"
  [runners.kubernetes]
    host = "https://45.67.34.123:4892"
    cert_file = "/etc/ssl/kubernetes/api.crt"
    key_file = "/etc/ssl/kubernetes/api.key"
    ca_file = "/etc/ssl/kubernetes/ca.crt"
    namespace = "gitlab"
    namespace_overwrite_allowed = "ci-.*"
    bearer_token_overwrite_allowed = true
    privileged = true
    cpu_limit = "1"
    memory_limit = "1Gi"
    service_cpu_limit = "1"
    service_memory_limit = "1Gi"
    helper_cpu_limit = "500m"
    helper_memory_limit = "100Mi"
    poll_interval = 5
    poll_timeout = 3600
    [runners.kubernetes.node_selector]
      gitlab = "true"
    [runners.kubernetes.node_tolerations]
      "node-role.kubernetes.io/master" = "NoSchedule"
      "custom.toleration=value" = "NoSchedule"
      "empty.value=" = "PreferNoSchedule"
      "onlyKey" = ""

Using volumes

As described earlier, volumes can be mounted in the build container. At this time hostPath, PVC, configMap, and secret volume types are supported. Users can configure any number of volumes for each of mentioned types.

Here is an example configuration:

concurrent = 4

[[runners]]
  # usual configuration
  executor = "kubernetes"
  [runners.kubernetes]
    [[runners.kubernetes.volumes.host_path]]
      name = "hostpath-1"
      mount_path = "/path/to/mount/point"
      read_only = true
      host_path = "/path/on/host"
    [[runners.kubernetes.volumes.host_path]]
      name = "hostpath-2"
      mount_path = "/path/to/mount/point_2"
      read_only = true
    [[runners.kubernetes.volumes.pvc]]
      name = "pvc-1"
      mount_path = "/path/to/mount/point1"
    [[runners.kubernetes.volumes.config_map]]
      name = "config-map-1"
      mount_path = "/path/to/directory"
      [runners.kubernetes.volumes.config_map.items]
        "key_1" = "relative/path/to/key_1_file"
        "key_2" = "key_2"
    [[runners.kubernetes.volumes.secret]]
      name = "secrets"
      mount_path = "/path/to/directory1"
      read_only = true
      [runners.kubernetes.volumes.secret.items]
        "secret_1" = "relative/path/to/secret_1_file"
    [[runners.kubernetes.volumes.empty_dir]]
      name = "empty-dir"
      mount_path = "/path/to/empty_dir"
      medium = "Memory"

Host Path volumes

HostPath volume configuration instructs Kubernetes to mount a specified host path inside of the container. The volume can be configured with following options:

PVC volumes

PVC volume configuration instructs Kubernetes to use a PersistentVolumeClaim that is defined in Kubernetes cluster and mount it inside of the container. The volume can be configured with following options:

ConfigMap volumes

ConfigMap volume configuration instructs Kubernetes to use a configMap that is defined in Kubernetes cluster and mount it inside of the container.

When using configMap volume, each key from selected configMap will be changed into a file stored inside of the selected mount path. By default all keys are present, configMap’s key is used as file’s name and value is stored as file’s content. The default behavior can be changed with items option.

items option is defining a mapping between key that should be used and path (relative to volume’s mount path) where configMap’s value should be saved. When using items option only selected keys will be added to the volumes and all other will be skipped.

Notice: If a non-existing key will be used then job will fail on Pod creation stage.

Secret volumes

Secret volume configuration instructs Kubernetes to use a secret that is defined in Kubernetes cluster and mount it inside of the container.

When using secret volume each key from selected secret will be changed into a file stored inside of the selected mount path. By default all keys are present, secret’s key is used as file’s name and value is stored as file’s content. The default behavior can be changed with items option.

items option is defining a mapping between key that should be used and path (relative to volume’s mount path) where secret’s value should be saved. When using items option only selected keys will be added to the volumes and all other will be skipped.

Notice: If a non-existing key will be used then job will fail on Pod creation stage.

Empty Dir volumes

emptyDir volume configuration instructs Kubernetes to mount an empty directory inside of the container.

Using Security Context

Pod security context configuration instructs executor to set a pod security policy on the build pod.

Assigning a security context to pods provides security to your Kubernetes cluster. For this to work you’ll need to provide a helper image that conforms to the policy you set here.

More about the helper image can be found here. Example of building your own helper image:

ARG tag
FROM gitlab/gitlab-runner-helper:${tag}
RUN addgroup -g 59417 -S nonroot && \
    adduser -u 59417 -S nonroot -G nonroot
WORKDIR /home/nonroot
USER 59417:59417

This example creates a user and group called nonroot and sets the image to run as that user.

Example of setting pod security context in your config.toml:

concurrent = %(concurrent)s
check_interval = 30
  [[runners]]
    name = "myRunner"
    url = "gitlab.example.com"
    executor = "kubernetes"
    [runners.kubernetes]
      helper_image = "gitlab-registy.example.com/helper:latest"
      [runners.kubernetes.pod_security_context]
        run_as_non_root = true
        run_as_user = 59417
        run_as_group = 59417
        fs_group = 59417

Using services

Define a list of services.

Service aliases are supported since GitLab Runner 12.9.

concurrent = 1
check_interval = 30
  [[runners]]
    name = "myRunner"
    url = "gitlab.example.com"
    executor = "kubernetes"
    [runners.kubernetes]
      helper_image = "gitlab-registy.example.com/helper:latest"
      [[runners.kubernetes.services]]
        name = "postgres:12-alpine"
        alias = "db1"
      [[runners.kubernetes.services]]
        name = "percona:latest"
        alias = "db2"

Using Docker in your builds

There are a couple of caveats when using Docker in your builds while running on a Kubernetes cluster. Most of these issues are already discussed in the Using Docker Build section of the GitLab CI documentation but it is worthwhile to revisit them here as you might run into some slightly different things when running this on your cluster.

Exposing /var/run/docker.sock

Exposing your host’s /var/run/docker.sock into your build container, using the runners.kubernetes.volumes.host_path option, brings the same risks with it as always. That node’s containers are accessible from the build container and depending if you are running builds in the same cluster as your production containers it might not be wise to do that.

Using docker:dind

Running the docker:dind also known as the docker-in-docker image is also possible but sadly needs the containers to be run in privileged mode. If you’re willing to take that risk other problems will arise that might not seem as straight forward at first glance. Because the Docker daemon is started as a service usually in your .gitlab-ci.yaml it will be run as a separate container in your Pod. Basically containers in Pods only share volumes assigned to them and an IP address by which they can reach each other using localhost. /var/run/docker.sock is not shared by the docker:dind container and the docker binary tries to use it by default.

To overwrite this and make the client use TCP to contact the Docker daemon, in the other container, be sure to include the environment variables of the build container:

• DOCKER_HOST=tcp://localhost:2375 for no TLS connection.
• DOCKER_HOST=tcp://localhost:2376 for TLS connection.

Make sure to configure those properly. As of Docker 19.03, TLS is enabled by default but it requires mapping certificates to your client. You can enable non-TLS connection for DIND or mount certificates as described in Use Docker In Docker Workflow with Docker executor

Not supplying Git

Do not try to use an image that doesn’t supply Git and add the GIT_STRATEGY=none environment variable for a job that you think doesn’t need to do a fetch or clone. Because Pods are ephemeral and do not keep state of previously run jobs your checked out code will not exist in both the build and the Docker service container. Errors you might run into are things like could not find git binary and the Docker service complaining that it cannot follow some symlinks into your build context because of the missing code.

Resource separation

In both the docker:dind and /var/run/docker.sock cases the Docker daemon has access to the underlying kernel of the host machine. This means that any limits that had been set in the Pod will not work when building Docker images. The Docker daemon will report the full capacity of the node regardless of the limits imposed on the Docker build containers spawned by Kubernetes.

One way to help minimize the exposure of the host’s kernel to any build container when running in privileged mode or by exposing /var/run/docker.sock is to use the node_selector option to set one or more labels that have to match a node before any containers are deployed to it. For example build containers may only run on nodes that are labeled with role=ci while running all other production services on other nodes. Further separation of build containers can be achieved using node taints. This will disallow other pods from scheduling on the same nodes as the build pods without extra configuration for the other pods.

Job execution

At the moment we are using kube exec to run the scripts, which relies on having a stable network connection between the Runner and the pod for the duration of the command. This leads to problems like Job marked as success midway. If you are experiencing this problem turn off the feature flag FF_USE_LEGACY_KUBERNETES_EXECUTION_STRATEGY to use kube attach for script execution, which is more stable.

We are rolling this out slowly and have plans to enable the kube attach behavior by default in future release, please follow #10341 for updates.

Using kaniko

Another approach for building Docker images inside a Kubernetes cluster is using kaniko. kaniko:

• Allows you to build images without privileged access.
• Works without the Docker daemon.

For more information, see Building images with kaniko and GitLab CI/CD.