Walter is a tiny deployment pipeline template.

• http://ainoya.io/walter
• http://walter-cd.net

Walter automates the deployment process of applications or servers from configuration to software deployment.

• Go 1.3 or greater
• Mercurial 2.9 or greater

You can build Walter with the following commands.

$ git clone git@github.com:walter-cd/walter.git
$ cd walter
$ ./build

We welcome any contributions through Github pull requests. When you make changes such as new features and the add the tests, please run test before throw the pull request. You can run test with the test.sh script.

$ sh test.sh

Walter has one configuration file, which specifies a set of tasks needed to build or deploy target application or service. More specifically, users specify the order of the tasks for the deployment. Each task is called as stage, and the flow is called pipeline in Walter.

The configuration format of Walter is Yaml. The yaml configuration file need to have one pipeline block, which has more than one stage element.

The following is a sample configuration of Walter.

pipeline:
  - name: command_stage_1
    type: command
    command: echo "hello, world"
  - name: command_stage_2
    type: command
    command: echo "hello, world, command_stage_2"
  - name: command_stage_3
    type: command
    command: echo "hello, world, command_stage_3"

As we see, the pipeline block has three stages and the stage type is command, each of which run echo command and has the stage name (such as command_stage_1). User can name arbitrary name of each stage. The commands are executed with the same order as the pipeline configuration.

Stage in pipeline has three elements, name, type and configurations. configuration elements are optional. The elements of configurations depend on the type. For example command_stage type has command configuration, which specify the shell command run in the stage. The following is the table on the type and the parameters.

Command stage executes one command. Users specify Command stage adding command in type.

The following is the parameter of Command stage.

Shell script stage executes specified shell script file. Users specify Shell script stage adding shell in type.

The following is the parameter of Shell script stage.

You can set child stages and run these stages in parallel like this.

pipeline:
  - name: parallel stages
    parallel:
      - name: parallel command 1
        type: command
        command: parallel command 1
      - name: parallel command 2
        type: command
        command: parallel command 2
      - name: parallel command 3
        type: command
        command: parallel command 3

In the above setting, parallel command 1, parallel command 2 and parallel command 3 are executed in parallel.

Walter provides the feature to import predefined stages in specified files. To import stages to pipeline configuration file, we use require block and add the list of file names into the block.

For example, the following example import the stages defined in conf/mystage.yml

require:
    - conf/mystages.yml

pipeline:
  - call: mypackage::hello
  - call: mypackage::goodbye

In the above setting, the stages ("mypacakge::hello" and "mypackage::goodbye") which are defined in "mystage.yml" are specified.

The files specified in pipeline configuration file need to have two blocks namespace and stages. In namespace, we add the package name, the package name is need to avoid collisions of the stage names in multiple required files. The stages block contains the list of stage definitions. We can define the stages same as the stages in pipeline configurations.

For example, the following configuration is the content of conf/mystages.yml imported from the above pipeline configuration file.

namespace: mypackage

stages:
  - def:
      name: hello
      command: echo "May I help you majesty!"
  - def:
      name: goodbye
      command: echo "Goobye majesty."

As we see that stages hello and goodbye are defined in the file.

Walter configuraiton can have one cleanup block; cleanup is another pipeline which needs to be executed after a pipeline has either failed or passed. In the cleanup block, we can add command or shell script stages. The below example create a log file in pipeline and then cleanup the log file in the cleaup steps.

pipeline:
  - name: start pipeline
    command: echo “pipeline” > log/log.txt
cleanup:
  - name: cleanup
    command:  rm log/*

Walter supports to submits the messages to messaging services.

To submit a message, users need to add a messenger block into the configuration file. The following is a sample of the yaml block with HipChat.

messenger:
  type: hipchat2
  base_url: BASE_URL
  room_id: ROOM_ID
  token: TOKEN
  from: USER_NAME

To report the full output of stage execution to the specified messenger service added with the above setting, users add report_full_output attribute with true into the stage they want to know the command outputs.

pipeline:
  - name: command_stage_1
    type: command
    command: echo "hello, world"
    report_full_output: true # If you put this option, messenger sends the command output "hello, world"
  - name: command_stage_2
    type: command
    command: echo "hello, world, command_stage_2"
    # By default, report_full_output is false

Walter supports HipChat API v1 and v2 as the messenger type.

To activate the report function, we need to specify the properties for messenger type. The needed properties are different in each messenger type.

Walter provides a coordination function to a project hosting service, GitHub. Specifically the service function provides two roles

• Check if a new commit or pull requests from the repository
• Run pipeline to the latest commit and pull requests when there are newer ones than the last update

To activate service coordination function, we add a "service" block to the Walter configuration file. "service" block contains several elements (type, token, repo, from, update).

service:
  type: github
  token: ADD_YOUR_KEY
  repo: YOUR_REPOSITORY_NAME
  from: YOUR_ACCOUNT_OR_GROUP_NAME
  update: UPDATE_FILE_NAME

The following shows the description of each element.

Users add environment variables in Walter configuration files. The names of environment variables are expanted into the the values of environment variables. Environment variables in the configuration files are valuable when we need to write the sensitive information such as tokens of messenger service or passwords of a external systems.

The following is the format of embedding of the environment variables.

$ENV_NAME

We write the envrionment variable into ENV_NAME. The following configuration file specify the GitHub Token by embedding the environment variable, GITHUB_TOKEN.

service:
  type: github
  token: $GITHUB_TOKEN
  repo: my-service-repository
  from: service-group
  update: .walter

Walter stores the results of preceding stages. The stages can make use of the results of finished stages using the three special variables (__OUT, __ERR, __COMBINED and __RESULT) in Walter configuration files.

• __OUT - output flushed to standard output
• __ERR - output flushed to standard error
• __COMBINED - combined output of stdout and stderr
• __RESULT - execution result (true or false)

The three variables are maps whose keys are stage names and the value are results of the stages. For example, we want the standard output result of the stage named "stage1", we write __OUT["stage1"].

The following is a sample configuration with a specical value.

pipeline:
  - name: stage_1
    command: echo "hello world"
  - name: stage_2
    command: echo __OUT["stage_1"]

Walter with the above configuraiton outputs "hello world" twice, since the second stage (stage_2) flushes the standard output result of first stage (stage_1).

Walter stage starts imidiately after the previous stage finish, but some stages need to wait for some action such as port is ready or file are created. wait_for feature supports the actions which need to be ready before the stages begin.

wait_for is defined as a property of stage.

pipeline:
  - name: launch solr
    command: bin/solr start
  - name: post data to solr index
    command: bin/post -d ~/tmp/foobar.js
    wait_for: host=localhost port=8983 state=ready

The wait_for property takes the key value pairs. Key has several variations. The value depends on the key type. The following table shows the supported key value pairs and the description.

There are seveal state values and possible state values are depend on the other key.