Mortar is a tool to easily handle a complex set of Kubernetes resources. Using kubectl apply -f some_folder/ is pretty straightforward for simple cases, but often, especially in CI/CD pipelines things get complex. Then on the otherhand, writing everything in Helm charts is way too complex.
While we were developing Kontena Pharos Kubernetes distro and tooling around it, we soon realized that we really want to manage sets of resources as a single unit. This thinking got even stronger while we were transitioning many of our production solutions to run on top of Kubernetes. As this is a common problem for all Kubernetes users, Mortar was born.
Mortar pre-baked binaries are available for OSX and Linux. You can download them from the releases page. Remember to put it in the path and change the executable bit on.
$ brew install kontena/mortar/mortar
Or to install the latest development version:
$ brew install --HEAD kontena/mortar/mortar
$ gem install kontena-mortar
To install bash/zsh auto-completions, use mortar install-completions after Mortar has been installed.
$ docker pull quay.io/kontena/mortar:latest
By default mortar looks for if file ~/.kube/config exists and uses it as the configuration. Configuration file path can be overridden with KUBECONFIG environment variable.
For CI/CD use mortar also understands following environment variables:
KUBE_SERVER: kubernetes api server address, for examplehttps://10.10.10.10:6443KUBE_TOKEN: service account token (base64 encoded)KUBE_CA: kubernetes CA certificate (base64 encoded)
$ mortar fire [options] <src-folder> <shot-name>
$ mortar yank [options] <shot-name>
$ mortar list [options]
$ mortar describe [options] <shot-name>
You can use mortar in CI/CD pipelines (like Drone) via quay.io/kontena/mortar:latest image.
Example config for Drone:
pipeline:
deploy:
image: quay.io/kontena/mortar:latest
secrets: [ kube_token, kube_ca, kube_server ]
commands:
- mortar fire k8s/ my-app
Namespace is mandatory to be set on the resources
Currently Mortar will not add any default namespaces into the resources it shoots. Therefore it is mandatory for the user to set the namespaces in all namespaced resources shot with Mortar. See this issue for details and track the fix.
Mortar manages a set of resources as a single unit, we call them shots. A shot can have as many resources as your application needs, there's no limit to that. Much like you'd do with kubectl apply -f my_dir/, but Mortar actually injects information into the resources it shoots into your Kubernetes cluster. This added information, labels and annotations, will be used later on by Mortar itself or can be used with kubectl too. This allows the management of many resources as a single application.
Most importantly, Mortar is able to use this information when re-shooting your applications. One of the most difficult parts when using plain kubectl apply ... approach is the fact that it's super easy to leave behind some lingering resources. Say you have some deployments and a service in your application, each defined in their own yaml file. Now you remove the service and re-apply with kubectl apply -f my_resources/. The service will live on in your cluster. With Mortar, you don't have to worry. With the extra labels and annotations Mortar injects into the resources, it's also able to automatically prune the "un-necessary" resources from the cluster. The automatic pruning is done with --prune option.
See basic example here.
One of the most powerful features of Mortar is it's ability to support overlays. An overlay is a variant of the set of resources you are managing. A variant might be for example the same application running on many different environments like production, test, QA an so on. A variant might also be a separate application "instance" for each customer. Or what ever the need is. Overlays in Mortar are inspired by kustomize.
Given a folder & file structure like:
echoservice-metallb/
├── echo-metal.yml
├── foo
│ └── pod.yml.erb
└── prod
├── pod.yml
└── svc.yml
where overlays (prod & foo) contain same resources as the base folder, mortar now merges all resources together. Merging is done in the order overlays are given in the command. Resources are considered to be the same resource if all of these match: kind, apiVersion, metadata.name and metadata.namespace.
If there are new resources in the overlay dirs, they are taken into the shot as-is.
You'd select overlays taken into processing with --overlay option.
Note: Overlays are taken in in the order defined, so make sure you define them in correct order.
The resources in the overlays do not have to be complete, it's enough that the "identifying" fields are the same.
See example of overlays here.
Mortar also support templating for the resource definitons. The templating language used is ERB. It's pretty simple templating language but yet powerful enough for Kubernetes resource templating.
Note: Mortar will process the resource definitions as ERB even if the filename does not have the .erb extension. This means that Ruby code in an innocent looking .yml file will be evaluated using the current user's access privileges on the local machine. Running untrusted code is dangerous and so is deploying untrusted manifests, make sure you know what you're deploying.
There are two ways to introduce variables into the templating.
See examples at examples/templates.
As for any process, environment variables are also available for Mortar during template processing.
kind: Pod
apiVersion: v1
metadata:
name: nginx
labels:
name: nginx
namespace: default
spec:
containers:
- name: nginx
image: nginx:<%= ENV["NGINX_VERSION"] || "latest" %>
ports:
- containerPort: 80Another option to use variables is via command-line options. Use mortar --var foo=bar my-app resources/.
Each of the variables defined will be available in the template via var.<variable name>.
kind: Pod
apiVersion: v1
metadata:
name: nginx
labels:
name: nginx
namespace: default
spec:
containers:
- name: nginx
image: nginx:latest
ports:
- containerPort: <%= port.number %>
name: <%= port.name %>You could shoot this resource with mortar --var port.name=some-port --var port.number=80 my-app resources/pod.yml.erb
It is also possible to pass variables, overlays and labels through a configuration file. As your templates complexity and the amount of variables used grows, it might be easier to manage the variables with an yaml configuration file. The config file has the following syntax:
variables:
ports:
- name: http
number: 80
- name: https
number: 443
overlays:
- foo
- barvariables, overlays and labels are optional.
For variables the hash is translated into a RecursiveOpenStruct object. What that means is that you can access each element with dotted path notation, just like the vars given through --var option. And of course arrays can be looped etc.. Check examples folder how to use variables effectively.
The configuration file can be given using -c option to mortar fire command. By default Mortar will look for shot.yml or shot.yaml files present in current working directory.
It's possible to set global labels for all resources in a shot via options or configuration file.
Use mortar --label foo=bar my-app resource to set label to all resources in a shot.
labels:
foo: bar
bar: bazBug reports and pull requests are welcome on GitHub at https://github.com/kontena/mortar.
Copyright (c) 2018 Kontena, Inc.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.