ocdeployer

A tool which wraps the OpenShift command line tools to enable repeatable automated deployment of OpenShift templates. Allows you to re-create environments based on templates more efficiently. Given a set of OpenShift templates, you can create a simple config that allows you to:

• Repeatedly deploy the same templates to different OpenShift projects
• Define the order in which they should deploy via 'stages'
• Optionally wait for resources to be "ready" before continuing on to the next stage:
  • DeploymentConfig resources to reach desired replica count
  • Build resources to succeed
  • StatefulSet resources to reach desired replica count
• Define which external 'images' should be imported to the project as ImageStreams
• Define which secrets your services rely on, and import them either from a local dir, or from another project in OpenShift
• Split component templates up into "service sets" and deploy all sets, or specific sets
• Define dependencies (for example: service set 'A' requires service set 'B')
• Create environment files, which define parameters that should be set at template processing time, so you can deploy the same templates to different environments
• Specify multiple environment files at deploy time and merge them
• Use OpenShift templating along with jinja2 templating
• Create custom pre-deploy/deploy/post-deploy scripts in python if more granular control is neeed
• Quickly scale the resource request/limit defined in your templates.
• Watch events in the namespace during deploy

REQUIRES OpenShift command line tools (the oc command)

You should log in to your project before deploying:

$ oc login https://api.myopenshift --token=*************

Getting Started

Details

A example ocdeployer project structure looks like the following:

├── env
│   ├── prod-env.yml
│   └── qa-env.yml
├── secrets
│   ├── mysql-secrets.yml
│   └── postgres-secrets.yml
└── templates
    ├── _cfg.yml
    ├── set1
    │   ├── _cfg.yml
    │   ├── env
    │   │   └── qa-env.yml
    │   ├── nginx.yml
    │   └── postgres.yml
    └── set2
        ├── _cfg.yml
        ├── custom
        │   └── deploy.py
        ├── mysql.yml
        └── ruby-app.yml

Components of the structure are explained below:

• secrets directory -- This is optional. Holds Openshift YAML files containing a Secret or List of Secret resources. Applications which require imported secrets can use the secrets in this directory.

• env directory -- You can create env files at the root level of the project, as well as in each service set (discussed below). You can specify multiple environments on the CLI at deploy time and the values will be merged.

• templates directory -- OpenShift YAML/JSON templates as well as a special config file (named _cfg.yml). Templates are typically split into into folders in this directory, called service sets. The base _cfg.yml defines the deploy order for all service sets, as well as any "global" secrets/images that should be imported that all services rely on.

• set1 and set2 are service set folders -- Each service set folder contains the following:

  • A _cfg.yml which configures deployment of that specific service set.
  • A directory named custom which houses custom deploy scripts. This is optional. Python scripts can be placed in here which can be run at pre-deploy/deploy/post-deploy stages for specific service sets.
  • A directory named env that contains an environment file or files. This is optional. Defines parameters that should be passed to the templates. Values defined in this env file are merged with values defined at the root level, with matching keys in the service set file taking precedence.

See the examples to get a better idea of how all of this configuration comes together and to get more details on what ocdeployer does when you run a deploy.

Installation and usage

Installation

You can install this package from pypi using:

$ pip install ocdeployer

Or, if you're installing from source, install into a virtual environment:

$ python3 -m venv .venv
$ . .venv/bin/activate
(venv) $ pip install -r requirements.txt

Usage

(venv) $ ocdeployer -h
Usage: ocdeployer [OPTIONS] COMMAND [ARGS]...

  Deploys components to a given cluster. NOTE: You need the openshift cli
  tool ('oc') installed and to login to your openshift cluster before
  running the tool.

Options:
  -h, --help  Show this message and exit.

Commands:
  deploy       Deploy to project
  list-routes  List routes currently in the project
  list-sets    List service sets available in template dir
  process      Process templates but do not deploy
  wipe         Delete everything from project

Deploy command

The deploy command is used to push your templates into OpenShift.

Use ocdeployer deploy -h for more details on parameters and options.

Process command

Use process to view the template data without actually deploying it. process has very similar options to deploy, but instead of pushing any configuration to OpenShift, it will simply parse the templates with jinja2 and OpenShift templating (i.e. running oc process), substitute in the given variables/project name/etc., and then either print out the resulting configuration to stdout or save the resulting processed files to a directory.

Wipe command

Use wipe to delete objects from a project. It essentially runs the following commands, deleting all objects or objects which have a specific label:

oc delete all [--all or --selector mylabel=myvalue]
oc delete configmap [--all or --selector mylabel=myvalue]
oc delete secret [--all or --selector mylabel=myvalue]
oc delete pvc [--all or --selector mylabel=myvalue]

List-routes command

Use list-routes to simply print the URLs for active routes in a project

List-sets command

Use list-sets to simply print the names of service sets that are available for deployment in your templates directory.

Template Configuration

The best way to explain how template configuration works is to describe the process for configuring a service set.

A guide to creating a service set

1. Create a new directory in the templates directory for your service set, e.g. "myservice"

2. Add your OpenShift YAML files for your services in this directory. The files should be OpenShift template files that contain all resources needed to get your service running (except for secrets, and image streams of external images, we'll talk about that shortly...). This would commonly be things like buildConfig, deploymentConfig, service, route, etc.

3. Create a '_cfg.yml' file in your directory. The contents of this config file are explained below:

   # (optional) requires
   #
   # Here you can list other service sets that need to be deployed before this one can.
   # Deployment will fail when processing this file if we see the required service set has
   # not yet been deployed in this run of ocdeployer.
   requires:
   - "myotherservice"
   
   # (optional) secrets
   #
   # Lists which secrets apps in this service set rely on; they will be validated or imported at run time.
   # A secret is only imported once per deploy run, so if other service sets rely on the same
   # secret it won't be imported again.
   secrets:
   - "mysecret"
   # You can also specify which service accounts a secret should be linked to
   - name: "othersecret"
     link: ["builder"]
   
   # (optional) custom_deploy_logic
   #
   # Indicates that there is a pre_deploy/post_deploy/deploy method defined for this
   # service set in the 'custom' folder that should be used
   custom_deploy_logic: true
   
   # (optional) post_deploy_timeout
   #
   # Indicates how long custom post_deploy logic should take before timeout (in seconds).
   # A null value can be handled differently depending on the post deploy logic,
   # but it is recommended that it means there is "no waiting" that will occur
   post_deploy_timeout: 300
   
   # (optional) images
   # Lists the image streams these services require to be imported into the destination namespace.
   #
   # key = the name or name:tag for the image stream. If no tag is specified, 'latest' is used.
   # value = the full image docker uri
   #
   # 'oc import-image <key> --from="<value>"' is run at deploy time.
   #
   # If it already exists, it will be re-imported (and therefore the image will be updated)
   images:
   - cp-kafka: "confluentinc/cp-kafka"
   - cp-zookeeper: "confluentinc/cp-zookeeper"
   - nginx-stable-openshift: "docker.io/mhuth/nginx-stable-openshift"
   - python-36-centos7: "centos/python-36-centos7"
   - postgresql-95-rhel7: "registry.access.redhat.com/rhscl/postgresql-95-rhel7"
   
   # (required) deploy_order
   #
   # Lists the order in which components in this service set should be deployed.
   deploy_order:
     # A stage deploys a group of components in sequentially and then waits for them to reach 'active'
     # in parallel. By default, at the end of each stage, we wait for:
     #   * DeploymentConfig's to be "active"
     #   * StatefulSet's to be "active"
     #   * BuildConfig's to succeed
     #
     # Setting 'wait' to false under the stage disables this behavior.
     #
     # Stages are processed after being sorted by name.
     stage0:
       wait: false
       components:
       - "zookeeper"
     stage1:
       # You can specify a wait timeout. By default, 300sec is used
       timeout: 600
       # 'components' lists the template files that should be deployed in this stage.
       # Their config is applied in openshift in the same order they are listed.
       components:
       - "kafka"
       - "inventory-db"
     stage2:
       components:
       - "insights-inventory"
       - "upload-service"
   

4. If you set custom_deploy_logic to True, read 'Custom Deploy Logic' below.

5. If you wish to define env vars for this service's templates, read 'Environment Files' below.

6. Add your service folder name to the base _cfg.yml in the templates directory. Remember that the deploy_order specifies the order in which each service set is deployed. So if your service set depends upon other services being deployed first, order it appropriately!

7. Run ocdeployer list-sets and you should see your new component listed as a deployable service set.

Custom Deploy Logic

By default, no pre_deploy/post_deploy is run, and the deploy logic is taken care of by the ocdeployer.deploy.deploy_components method. So, unless you are doing something complicated and require additional "python scripting" to handle your (pre/post) deploy logic, you don't need to worry about custom logic (and it's quite rare that you'd want to re-write the main deploy logic itself)

But let's say you want to perform some tasks prior to deploying your components, or after deploying your components. Custom scripts can be useful for this.

You can set custom_deploy_logic in your service set's _cfg.yml to true. You then have two options for defining a custom deploy script:

• You can define a single deploy.py in the root custom directory of your project. This script will apply to all service sets as long as they have custom_deploy_logic set to true.
• You can create a script called deploy.py in the custom dir of your service set. This script will apply only to that service set.

If a project contains a deploy.py in the root custom directory as well as in the service set's custom directory -- the service set deploy script takes precedence.

NOTE

In previous versions of ocdeployer<4.0, the custom dir was housed in the root folder of the project, and the deploy file inserted in there needed to match the name of your service set, e.g. deploy_myservice.py. For backward compatibility, this method of defining the service set deploy scripts is still supported.

Inside this script you can define 3 methods:

def pre_deploy(project_name, template_dir, variables_per_component):

• project_name: string, name of project being deployed to
• template_dir: string, the full path to the directory of the service set's templates which are being deployed
• variables_per_component: dict, keys are each component name in your service set, values are another dict consisting of the variables parsed from the env.yml file.

def deploy(project_name, template_dir, components, variables_per_component, wait, timeout, resources_scale_factor, label):

• project_name: string, name of project being deployed to
• template_dir: string, the full path to the directory of the service set's templates which are being deployed
• components: list of strings, the component names from your service set that are being deployed
• variables_per_component: dict, keys are each component name in your service set, values are another dict consisting of the variables parsed from the env.yml file.
• wait: boolean, used to determine whether the deploy logic should wait for things such as DeploymentConfig and BuildConfig to "finish" (go 'active', or build with success, respectively)
• timeout: int, how long to wait for before timing out
• resources_scale_factor: float, the value passed in to --scale-resources when running ocdeployer
• label: string, the label attached to each object in Open Shift at deploy time

def post_deploy(processed_templates, project_name, template_dir, variables_per_component):

• processed_templates: dict with containing the processed template info for each component that was deployed -- keys: template name, vals: an instance of ocdeployer.templates.Template
• project_name: string, name of project being deployed to
• template_dir: string, the full path to the directory of the service set's templates which are being deployed
• variables_per_component: dict, keys are each component name in your service set, values are another dict consisting of the variables parsed from the env.yml file.

Much of the code in ocdeployer.common may be useful to you as you write custom deploy logic (such as the oc method used to run oc commands).

Custom Deploy Example 1

Let's say that after you deploy your components, you want to trigger a build on any build configurations you pushed (actually, this already happens after each stage by default as long as wait is not false on the stage in your service set _cfg.yml -- but play along for the sake of this example).

You could define a post-deploy method that looks like this:

from ocdeployer.utils import oc, wait_for_ready_threaded

log = logging.getLogger(__name__)

def post_deploy(**kwargs):
    build_config_names = []
    for _, template in kwargs.get("processed_templates", {}).items():
        # Get the name of all build configs that were deployed
        # Remember, we are looking at the processed template info
        # We're looking at the template AFTER variable substitution occurred.
        build_config_names.extend(template.get_processed_names_for_restype("bc"))

    objs_to_wait_for = []
    for bc_name in build_config_names:
        oc("start-build", bc_name, exit_on_err=False)
        objs_to_wait_for.append(("bc", bc_name))
    else  
        log.warning("No build configs were deployed, nothing to do")

    # Wait for all builds to reach 'completed' state:
    if objs_to_wait_for:
        wait_for_ready_threaded(objs_to_wait_for)

Custom Deploy Example 2

Let's say that when any ConfigMap resources in your project update, you want to trigger a new rollout of a deployment. You could do that by tracking the state of the ConfigMap before deploying and comparing it to the state after.

from ocdeployer.utils import oc, get_json, wait_for_ready

log = logging.getLogger(__name__)
old_config_map_data = {}

def pre_deploy(**kwargs):
    old_config_map_data = get_json("configmap", "MyConfigMap")['data']


def post_deploy(**kwargs):
    new_config_map_data = get_json("configmap", "MyConfigMap")['data']
    if new_config_map_data != old_config_map_data:
        oc("rollout", "dc/MyDeployment")
        wait_for_ready("dc", "MyDeployment")

Images

You can use the images section in a _cfg.yml to instruct ocdeployer to import images and configure an ImageStream for them.

ocdeployer will run the following for each required image at deploy time.

oc import-image <istag> --from="<from>" --scheduled=true --confirm

If the image already exists, it will be re-imported (and therefore the image will be updated).

If the ImageStreamTag already exists, but the "from" URI has been updated, ocdeployer will run the following to re-configure the tag on the ImageStream:

oc tag --scheduled=true --source=docker <from> <istag>

If the same ImageStreamTag is defined more than once (whether in the same _cfg.yml or in various _cfg.yml files), it is not imported repeatedly. For this reason, ImageStreamTag's need to be uniquely named across service sets if they are intended to be deployed into the same namespace.

The images section can be defined in two ways in _cfg.yml:

Short format

List the desired ImageStreamTag as the key, and the image's "from" URI as the value.

If no tag is specified, it is assumed that "latest" is the desired tag.

images:
- "cp-kafka:sometag": "confluentinc/cp-kafka"

Long format (ocdeployer>=v4.3.0)

• istag -- the desired ImageStreamTag
• from -- the image's external "from" URI
• envs -- lists envs this image should be imported for. The image will only be imported when ocdeployer is run with --env matching the envs given. If no envs are given, it is assumed it should be loaded in all envs.

images:
- istag: "cp-kafka:sometag"
  from: "confluentinc/cp-kafka"
  envs: ["qa", "prod"]

Secrets

By default, ocdeployer will simply validate that the secrets specified in _cfg.yml already exist in your destination namespace.

You can run ocdeployer with --secrets-src-project to import the required secrets into your project from a different project in OpenShift, or use --secrets-local-dir to load the required secrets from OpenShift config files stored in a local directory.

If you set --secrets-src-project to be the same as the destination namespace, this effectively causes ocdeployer to simply validate that the secret is present in that namespace.

Any .yaml/.json files you place in the secrets-local-dir will be parsed and secrets will be pulled out of them and imported. The files can contain a single secret (kind: Secret) OR a list of resources (kind: List)

An example secret .yaml file:

apiVersion: v1
data:
    ssh-privatekey: <your key here>
kind: Secret
metadata:
    creationTimestamp: null
    name: my_secret
type: kubernetes.io/ssh-auth

Secrets can be specified in _cfg.yml in two ways:

Short format

List just the name:

secrets:
- "name of secret"

Long format (ocdeployer>=v4.3.0)

List the name as well as:

• link -- lists service accounts the secret should be linked to using oc secrets link
• envs -- lists envs this secret should be imported for. The secret will only be imported when ocdeployer is run with --env matching the envs given. If no envs are given, it is assumed it should be loaded in all envs.

secrets:
- name: "name of secret"
  link: ["account1", "account2"]
  envs: ["qa", "prod"]

Note that any links existing for a secret will not be removed at deploy time -- ocdeployer will only add new links.

How do I export secrets from a project to use later with --secrets-local-dir?

You can export all with:

$ oc export secrets -o yaml > /tmp/secrets/secrets.yaml

Or export a single secret object:

$ oc export secret mysecret -o yaml > /tmp/secrets/mysecret.yaml

To use the secrets files in your next project deploy:

(venv) $ oc login https://my.openshift --token=*************
(venv) $ ocdeployer deploy -a --secrets-local-dir /tmp/secrets/ myproject

Environment files

By default, the following parameters are passed to templates by ocdeployer at deploy time:

• 'NAMESPACE' corresponds to the project name selected on the CLI.
• 'SECRETS_PROJECT' corresponds to the secrets-src-project selected on the CLI

You can define "environment" files in two places with more customized variable information.

Root environment files

In the root of your project, a directory called env can house environment yaml files which set variables at the root level of your project. Let's say we create a file called env/myenv.yaml which looks like this:

global:
  # Values defined outside of the "parameters" section are intended to be evaluated during jinja2 processing
  some_var: false
  parameters:
    # Values defined as "parameters" are evaluated by 'oc process' as OpenShift template parameters
    VAR1: "applies to all components"
    VAR2: "also applies to all components"

set1:
  parameters:
    VAR2: "this overrides global VAR2 for only components in the 'set1' set"
    VAR3: "VAR3 applies to all components within the 'set1' service set"

set1/db:
  parameters:
    VAR2: "this overrides global VAR2, and 'set1' VAR2, for only the 'db' component"
    VAR4: "VAR4 only applies to 'db' component"
    # Using keyword {prompt} will cause ocdeployer to prompt for this variable's value at runtime.
    VAR5: "{prompt}"

This allows you to define your variables at a global level, at a "per service-set" level, or at a "per-component within a service-set" level. You can override variables with the same name at the "more granular levels" as well. If an OpenShift template does not have a variable defined in its "parameters" section, then that variable will be skipped at processing time. This allows you to define variables at a global level, but not necessarily have to define each variable as a parameter in every single template.

Service set environment files

You can also define variables inside an env directory for each service set. Let's say we create a file at templates/set1/env/myenv.yaml which looks like this:

global:
  # Values defined here apply to all components of service set 'set1'
  parameters:
    VAR1: "overrides the VAR1 set globally at root level"

db:
  parameters:
    VAR5: "overrides the VAR5 set on db component at root level"

The variable merging and variable overriding process works the same as for the root level env file. The service-set env file is merged with the root env file at processing time. But in addition, the values defined at the service-set level take precedence over any values defined at the root level.

Selecting environment

Select your environment at runtime with the -e or --env command-line option, e.g.:

(venv) $ ocdeployer deploy -s myset -e myenv myproject

NOTE

In ocdeployer<v4.0, there was no support for env files within the service set, so only root environment files were used and the -e/--env-file argument was used to point to specific YAML file paths. This is still supported for backward compatibility.

Overriding deploy config via environment file

The component key name _cfg is reserved in environment files. Data listed under this key can be structured in the same way as a _cfg.yml file, allowing you to add to or override values of the deloy config for different environments.

• A _cfg section defined in a base env file will be merged with values set in the base _cfg.
• A _cfg section defined in a service set's env file will be merged with values set in the _cfg for that service set.

As an example, let's say that you want to import an image only when deploying to the "qa" environment.

Your base config may look like this:

deploy_order:
  stage0:
    components: ["set1", "set2", "set3"]

Your service set config for set1 may look like this:

images:
- image1: repo/image1:latest

deploy_order:
  stage0:
    components: ["component1", "component2"]

You could then define templates/set1/env/qa.yml with this setting:

_cfg:
  images:
  - image1: repo/image1:other-tag
  - image2: repo/image2:latest

This will cause the following to occur only when --env qa is selected at deploy time:

• the source image of the image1 ImageStreamTag to be overridden to map to a new external tag (other-tag instead of latest)
• image2 to be imported

Using multiple environment files

You can define multiple environments and merge them at deploy time. Example:

env1.yaml

global:
  my_value: true
  my_other_value: false

env2.yaml

global:
  my_value: false
  my_other_value: true

Running the following command:

(venv) $ ocdeployer deploy -s myset -e env1 -e env2 myproject

Results in env1.yaml and env2.yaml being merged. Since env1 is listed FIRST in the list, any matching parameter entries in this file will override those of env2. The result is a values file which looks like:

global:
  my_value: true
  my_other_value: true

Common usage

List the service sets available for you to deploy:

(venv) $ ocdeployer list-sets
Available service sets: ['platform', 'advisor', 'engine', 'vulnerability']

Example to deploy platform/engine service sets using "prod" env, and import secrets from "mysecretsproject":

(venv) $ ocdeployer deploy -s platform,engine -e prod --secrets-src-project mysecretsproject mynewproject

You can scale the cpu/memory requests/limits for all your resources using the --scale-resources flag:

(venv) $ ocdeployer deploy -s platform --scale-resources 0.5 mynewproject

This will multiply any configured resource requests/limits in your template by the desired factor. If you haven't configured a request/limit, it will not be scaled. If you scale by 0, the resource configuration will be entirely removed from all items, causing your default Kubernetes limit ranges to kick in.

Delete everything (and that means pretty much everything, so be careful) from your project with:

(venv) $ ocdeployer wipe <openshift project name>

You can also delete everything matching a specific label:

(venv) $ ocdeployer wipe -l mylabel=myvalue <openshift project name>

Known issues/needed improvements

• The scripts currently check to ensure deployments have moved to 'active' before exiting, however, they do not remediate any "hanging" or "stuck" builds/deployments.
• There is currently no way to "enforce" configuration -- i.e. delete stuff that isn't listed in the templates.