cr 📂

The concurrent runner

Overview

cr is a job executor concerned with achieving the highest parallel execution possible.

Given a definition of jobs and their dependencies, it builds a graph that outlines the execution plan of these jobs.

For instance, consider the following plan:

Jobs:
  - Id: 'SayFoo'
    Run: 'echo foo'

  - Id: 'SayBaz'
    Run: 'echo baz'
    DependsOn: [ 'SayFoo' ]

  - Id: 'SayCaz'
    Run: 'echo caz'
    DependsOn: [ 'SayFoo' ]

This plan states that we have 3 jobs to be executed: SayFoo, SayBaz and SayCaz but the last two jobs must only be executed after the first one and in case it succeeds.

To visualize the execution plan we can run it with --graph, which validates the plan and prints out a dot digraph.

# Execute CR telling it where the execution
# plan is (execution.yaml) and that it should
# just print the graph and exit.
cr --file ./execution.yaml  --graph

digraph {
	compound = "true"
	newrank = "true"
	subgraph "root" {
		"[root] SayFoo" -> "[root] SayBaz"
		"[root] SayFoo" -> "[root] SayCaz"
		"[root] _root" -> "[root] SayFoo"
	}
}

# If we pipe this to `dot` and than take the output
# of `dot` we can see the visual representation of the 
# digraph.

cr --file ./examples/hello-world.yaml --graph \
        | dot -Tpng > ./assets/hello-world.graph.png

Spec

---
# Configurations that control the runtime environment.
# These are configurations that can be specified via
# the `cr` CLI (cli takes precedence).
Runtime:
  LogDirectory: '/tmp'  # base directory to use to store log files
  Stdout: false         # whether all logs should also go to stdout     
  Directory: './'       # default directory to be used as CWD


# Map of environment variables to include in every job 
# execution.
# This can be be overriden by job-specific environments
Env:
  FOO: 'BAR'


# Jobs is a list of `Job` objects.
# Each job can have its properties templated
# using results of other jobs, even if they
# depend on the result of a job execution.
Jobs: 
  - Id: MyJob           # name of the job being executed.
    Run: 'echo test'    # command to run
    Directory: '/tmp'   # directory to use as cwd in the execution
    CaptureOutput: true # whether the output of the task should be stored in `.Output` variable
    Env:                # Variables to merge into the environment of the command
      FOO: 'BAR'
    DependsOn:          # List of strings specifying jobs that should be executed before this 
      - 'AnotherJob'    # job and that must exit succesfully.
    LogFilepath: '/log' # Path to the file where the logs of this execution should be stored.
                        # By default they're stored under `/tmp/<NameOfTheJob>`.