kubedev

DevOps command line tool that standardizes workflows for Microservices in Kubernetes for teams:

> Develop, Build, Secure, Deploy.

It builds on well-known and field-proven tools:

• docker
• tilt
• helm
• kind
• Security audits:
  • Fairwinds Polaris
• CI providers:
  • GitLab

kubedev Principles

• kubedev wants to help you quickly and easily build microservices that are independent, but at the same time follow a common pattern in regards to building, documenting and deploying. This makes it easier to add new services, and to onboard new developers.
• kubedev aims to be a thin wrapper around the commands it builds on, and just wants to make it easier for teams to call them appropriately.
• kubedev always prints the commands that it executes, so that you know what is going on.
• kubedev heavily relies on environment variables for service configuration.
• kubedev helps you build your services "secure by default".

Current state of development

kubedev is in early development and used internally at Gira.

Synopsis

kubedev commands are based on the definitions found in kubedev.json, which include the minimum necessary information that is required to execute common cloud-dev related tasks.

A kubedev.json describes an "Service", which in turn can contain "Apps" that may be deployments or cronjobs.

Schema of kubedev.json:

{
    "name": "myservice",
    "description": "My fancy service 🎆",
    "imagePullSecrets": "foo-creds", # Your docker registry auth credentials
    "imageRegistry": "foo-registry", # Your docker registry
    "helmReleaseName": "myservice-v1", # An optional name of the helm release. If not specified, 'name' will be used.
    "securityChecks": {
      "polaris": { # Optional, specify a custom polaris configuration
        "configFile": "polaris-config-cli.yaml", # Local file name of the polaris config
        "configDownload": { # Optional, download the polaris config to the local file before running the audit
          "url": "https://url-to-your-polaris-config",
          "headers": { # Optional, specify headers added to the GET request to download the polaris config
            "Authentication": "Bearer ${YOUR_BEARER_TOKEN}"
          }
        }
      },
      "variables": {
        "MYDEPLOY_FLASK_ENV": "development" # You can override variables here that are set when calling `helm template` to pipe into the security checker
      }
    },
    "required-envs": {
      "MYSERVICE_ENV": {
        "documentation": "Describe MYSERVICE_ENV here, so that other devs on your team know how to set them in their own env",
        "container": true, # Use this environment variable when running containers
        "build": true # Use this environment variable for building the container
      }
    },
    "deployments": {
        "mydeploy": { # An App `mydeploy' of type deployment
            "usedFrameworks": ["python", "pipenv", "npm", "vue"], # Not implemented, yet. usedFrameworks are used to e.g. fill in Tiltfile live_update, ignore, etc.
            "ports": {
              "https": {
                  "container": "8081", # This is the port that your actual dockerized service is bound to
                  "service": "8082",   # This is the port that the Kubernetes service serves on. Will be redirected to the container-port of the pods.
                  "dev": "8083" # This is the port used for local development by either `tilt` or `kubedev run`. Will be available on localhost when using `tilt up` or `kubedev run`.
              }
            },
            "volumes": {
              "dev": {
                "other_path": "/short/path" # Shorthand to mount a path read-write,
                "host_path": {
                  "path": "/container/path", # Mount local directories to container directories when running via `kubedev run`
                  "readOnly": true # mount them read-only
                },
                "inline_content.txt": {
                  "content": "Hello, World!", # Mount fixed, plain-text content to a file at 'path'
                  "path": "/container/path",
                  "readOnly": true # mount file read-only
                },
                "inline_content_base64.txt": {
                  "path": "/container/path", # Mount fixed, base64-encoded content to a file at 'path'
                  "base64": "SGVsbG8sIFdvcmxkIQ==",
                  "readOnly": true # mount file read-only
                }
              }
            },
            "required-envs": {
                "MYDEPLOY_FLASK_ENV": {
                    "documentation": "..."
                },
                "MYDEPLOY_COMPLEX": {
                  "documentation": "This is a variable with content that can not be passed to the helm chart on the command line and must be base64-encoded.",
                  "transform": "base64"
                }
            },
            # Defines the system test for this app. System tests must be defined in the directory ./systemTest// and must include a Dockerfile:
            "systemTest": {
                "variables": {
                  # Defines variables that are passed as environment variables into the test container and all services:
                  "TEST_GLOBAL_VAR": "value"
                },
                "testContainer": {
                    "volumes": {
                        "local_file": "/tmp/container_file" # You can mount volumes into the testContainer
                    },
                    "variables": {
                        # Defines environment variables that are passed to the system-test container
                        "TEST_HOSTNAME": "myservice-test"
                    },
                    "buildArgs": {
                        # Defines build arguments for the system test container
                        "TEST_BUILD_X": "X"
                  }
                },
                # Services run in the background, in an isolated docker network, while the system test container is executed
                "services": {
                    # Services can reference deployments defined in this kubedev.json and will use the latest available image tag,
                    # as generated by e.g. `kubedev build`.
                    "{mydeploy}": {
                        "hostname": "myservice-test", # The system test container can access this service by this hostname
                        "ports": [8081], # Ports that are accessible from other containers
                        "cmd": ["--some", "commandline", "--args", "${SHELL_EXPANDED_VARIABLE}"], # Arguments passed to 'docker run...' after the image-name
                        "variables": {
                            # Environment variables that are passed to this service.
                            # All required-envs for this deployment are passed, too, but can be overwritten here:
                            "MYSERVICE_ENV": "test-value"
                        }
                    },
                    # Services can use public images or images from your private repository, too:
                    "postgres:13": {
                        "hostname": "postgres-test",
                        "ports": [5432], # Ports that are accessible from other containers
                        "variables": {
                            "POSTGRES_USER": "tempuser",
                            "POSTGRES_PASSWORD": "correct horse battery staple"
                        }
                    }
                }
            }
        }
    },
    "cronjobs": {
        "myjob": {
            "volumes": {
              "dev": {
                  "job_files/": "/tmp/job_files/"
              }
            },
            "required-envs": {
                "MYJOB_VAR": {
                    "documentation": "..."
                }
            },
            "systemTest": {
                # See the detailed documentation on system tests below.
                "services": {
                    # Services can use public images or images from your private repository, too:
                    "postgres:13": {
                        "hostname": "postgres-test",
                        "ports": [5432], # Ports that are accessible from other containers
                        "variables": {
                            "POSTGRES_USER": "tempuser",
                            "POSTGRES_PASSWORD": "correct horse battery staple"
                        }
                    }
                }
            }
        }
    },
    "generic": {
        "something": {
            "required-envs": {
              # Put some required-envs here that are not used by something that is built as a container
          }
      }
    }
}

Naming conventions

kubedev will work with artifacts that follow certain naming conventions that are built from the \ (top level "name"), the \ (the "name" inside of "deployments" and "cronjobs") and a tag, which will be used as the image tag.

Artifact	Naming Convention
helm chart name	The helm chart is generated with the name ``.
helm release name	The release name is either directly specified using helmReleaseName in kubedev.json, or is the same as the helm chart name if helmReleaseName is not specified.
kubernetes labels	All kubernetes definitions include the labels kubedev-deployment or kubedev-cronjob () and kubedev-app .
image name	The image name is built using /-, except when \ is the same as , in which case it is collapsed to just /
image tag	The tag is either built using "${CI_COMMIT_SHORT_SHA}_${CI_COMMIT_REF_NAME}", if both these environment variables are set, or to none otherwise.

Automatic docker login

When kubedev needs to access docker registries, it writes the content of ${DOCKER_AUTH_CONFIG} to the file ${HOME}/.docker/config.json, if:

• ~/.docker/config.json does not exist.
• The environment variable ${HOME} is set.
• The environment variable ${CI} is set.
• The environment variable ${DOCKER_AUTH_CONFIG} is set.

Environment Variable Transformation

Environment Variables are passed to helm and docker by shell expansion. This has some limitations of values that are not "shell-safe", such as when they contain double-quotes or special characters. To make it safe and possible to pass these values, kubedev provides a few transformations.

A transformation will take the content of the variable from the environment where kubedev is called, and pass it in a transformed way into helm --set and docker --build-arg / --env.

Available transformations are:

Transformation	Description
base64	Base64 encodes the value

You can enable a transformation by setting the attribute transform in required-envs to the desired transformation, e.g. base64.

kubedev generate

Creates artifacts that kick-starts your microservice development.

The following files are created for by - with secure defaults:

• Tiltfile
• .gitlab-ci.yml
• helm-chart/ with all required kubernetes ressources
• /Dockerfile

See Naming Conventions.

The following "usedFrameworks" are supported by kubedev generate.

pipenv

When "pipenv" is included in the "usedFrameworks" by either an app or globally, the following files are generated inside the 's sub-directory:

File	Description
Dockerfile	Includes pipenv-specific instructions
app.py	An empty python script
Pipfile	A Pipfile using python 3.8. No packages or dev-packages are added.
Pipfile.lock	The result of locking the empty Pipfile.

kubedev check

Reads kubedev.json and checks whether all environment variables from the configuration is set in the current environment. It prints missing variables, including it's documentation.

kubdev check's scope can be narrowed down by optional command line arguments:

• kubedev check without additional arguments checks whether all environment variables used for build and runtime exist, and prints the missing environment variables.
• kubedev check [cmd] checks only the environment variables required for the specified task. E.g. kubedev check build will check for the existance of environment variables where "build" is set to true (or "build" is not specified, which defaults to true).
• kuebdev check [cmd] [app2] [app2] ... checks only the environment variables that are required for the specified task (e.g. kubedev check build) for the given apps (deployments, cronjobs or generic).

kubedev audit

Audits the k8s specification using Fairwind's Polaris.

A custom configuration file can be specified via the configuration options securityChecks.polaris, see the example.

Possible configuration options:

JSON field	Description
securityChecks.variables	A dictionary with variables and values that are used to override the current environment's values. You can use these to set specific values when executing helm template to generate the Kubernetes ressources that are checked by polaris.
securityChecks.polaris.configFile	A path to a local configuration file that is passed to polaris. The file must exist or configDownload must be specified. Environment variables will be expanded.
securityChecks.polaris.configDownload	An optional object that can be used to specify a download path where kubedev audit will fetch the polaris configuration from.
securityChecks.polaris.configDownload.url	The URL to the polaris config. Environment variables will be expanded. When the object configDownload exists, the field url is mandatory.
securityChecks.polaris.configDownload.headers	An optional dictionary that can contain custom headers that will be passed to the polaris config download. Both the header names and values can include environment variables.

Note: The polaris executable needs to be available in your $PATH.

kubedev build \

Runs docker build for \ with all docker build args as defined in kubedev.json.

See Naming Conventions.

This commands writes the content of ${DOCKER_AUTH_CONFIG} to the file ${HOME}/.docker/config.json, if:

• ~/.docker/config.json does not exist.
• The environment variable ${CI} is set.
• The environment variable {DOCKER_AUTH_CONFIG} is set.

kubedev build is used inside the CI/CD build jobs generated by kubedev generate and internally by the kubedev run command.

kubedev push \

Runs docker push for .

See Naming Conventions.

See Automatic docker login

kubedev push is used inside the CI/CD build jobs generated by kubedev generate.

kubedev run \

Builds an image using kubedev build with a random tag and then runs it.

The following parameters are passed to docker run, some of them can be configured in kubedev.json:

• --interactive and --rm are always passed
• If kubedev is called from a terminal, --tty is passed.
• required-envs are forwarded to the container.
• All ports.\.dev will be forwarded to ports.\.container
• volumes.dev are passed to the container. kubedev will auto-detect a WSL + Docker Desktop environment and convert the source path to a Windows path using wsl-path -aw.

kubedev deploy

Reads a kube config from the env var $KUBEDEV_KUBECONFIG (required), writes it to ** and optionally a context from $KUBEDEV_KUBECONTEXT and then runs helm upgrade --install --kube-config " with appropriate arguments and all required env vars from kubedev.json.

See Naming Conventions.

The generated image tag is passed to the template using the environment variable KUBEDEV_TAG.

kubedev deploy is used inside the CI/CD build jobs generated by kubedev generate.

kubedev template

Basically runs helm template with appropriate arguments and env vars from kubedev.json.

See Naming Conventions.

The generated image tag is passed to the template using the environment variable KUBEDEV_TAG.

kubedev template is used inside the Tiltfile generated by kubedev generate.

kubedev system-test \

System Tests for Deployments

If \ is a deployment, the system-test is run in "deployment"-mode, which is different and much quicker than the "cronjob"-mode (see below).

In "deployment"-mode, a system-test container is run against defined services, which can either be services defined in kubedev.json, or services pulled from a registry such as a database or a service defined in another repository. The system-test container and the services run directly in docker and not in Kubernetes. Note that local kubedev invocations behave differently from its invocations being inside a CI pipeline. When running kubedev system-test locally, all container images necessary for a given system-test are built. When running in CI, only the system-test images themselves are built. Deployments & CronJob images are expected to be pullable. Hence, in CI system-tests may never preceed their corresponding build-push stage.

See Automatic docker login

kubedev system-test's behaviour is defined in a "systemTest" sub-element of the deployment definition. See the thorough example at the beginning of the README.md.

The following configuration options are available:

Configuration element	Description	Mandatory
systemTest.variables	Defines global variables that are passed as environment variables into the services (if defined) and the system test container	No
systemTest.testContainer.variables	Defines variables that are passed as environment variables into the system test container, but not the services	No
systemTest.testContainer.buildArgs	Defines build args that are used when building the system test container	No
systemTest.testContainer.services	Defines services that are run in the background when running the system test container. Use the syntax {\ to reference deployments that are defined in this kubedev.json, and the image name will be built automatically according to the same rules as kubedev build. The app's volumes.dev are passed to the container as kubedev run does.	No
systemTest.testContainer.services[...].hostname	This service will be available by this hostname from the system test container	Yes
systemTest.testContainer.services[...].ports	Defines ports that are published from this service	No
systemTest.testContainer.services[...].cmd	A list of command line arguments passed to the docker run... command.	No
systemTest.testContainer.services[...].variables	Defines additional environment variables that are passed to this service. When this service references a kubedev deployment, additionally all required-envs for this deployment are passed into the service. These values can be overwritten using this variables.	No

The schematic flow when running the system-test is as follows:

1. Build the system test container from ./systemTest/\/.
2. Create a temporary docker network.
3. Try to remove left-over services from previous run.
4. Run all defined services in the background.
5. Wait for all services to become ready.
6. Run the system test container in the foreground.
7. Remove the services containers.
8. Remove the temporary docker network.

System Tests for CronJobs

Introduction

If \ is a cronjob, the sytsem-test is run in "cronjob"-mode, which spins up a temporary Kubernetes cluster using kind.

You can pass --reuse-cluster to speed up successive runs of system-tests. On the first run, if the kind cluster does not exist, it is created. On all successive runs this cluster is re-used, which speeds up test execution time. The details in the section "Flow" below.

See Automatic docker login

kubedev system-test's behaviour is defined in a "systemTest" sub-element of the cronjob definition.

Configuration

The following configuration options are available:

Configuration element	Description	Mandatory
systemTest.clusterInitChart	Defines a path to a helm chart that is deployed after cluster creation. This helm chart can contain prerequisites for your service	No
systemTest.variables	Defines global variables that are passed as environment variables into the services (if defined) and the system test container	No
systemTest.testContainer.variables	Defines variables that are passed as environment variables into the system test container, but not the services	No
systemTest.testContainer.buildArgs	Defines build args that are used when building the system test container	No
systemTest.testContainer.services	Defines services that are run outside of the cluster when running the system test container. This can not reference apps from the kubedev.json, because the whole service is deployed to the cluster and hence all apps run, anyways.	No
systemTest.testContainer.services[...].hostname	This service will be available by this hostname from the system test container	Yes
systemTest.testContainer.services[...].ports	Defines ports that are published from this service	No
systemTest.testContainer.services[...].cmd	A list of command line arguments passed to the docker run... command.	No
systemTest.testContainer.services[...].variables	Defines additional environment variables that are passed to this service.	No

Flow

The schematic flow when running the system-test is as follows:

1. Build the system test container from ./systemTest/\/.
2. Build and push all apps from kubedev.json.
3. Spin up a new kind cluster. (This step is skipped when re-using a cluster.)
4. Run a special service that can be used by the system-tests to easily start the CronJob. See the API description below.
5. Cluster initialization: , install tiller using helm init. (This step is skipped when re-using a cluster.)

• Create a tiller service-account with cluster-admin permissions. (This step is skipped when re-using a cluster.)
• Run helm init to install tiller. (This step is skipped when re-using a cluster.)
• Create the secret defined by imagePullSecrets with the content `.dockerconfigjson: