Kube Scheduler

A kubernetes utility to run docker images on a cron-like schedule.

Why We Made This

We love (and use!) kubernetes, docker and microservices so instead of trying to shove new schedulable jobs into our existing Resque architecture we built this utility to schedule and run docker images inside of our kubernetes cluster.

To Use

The scheduler requires a kubernetes cluster. If you have one set up, great keep reading! If not there are guides available to walk you through setting up a local or hosted solution (minikube is a really simple way to get started).

Once your cluster is configured you need a job/task to run. Our workflow looks like this:

1. Write the code for our job/task.
2. Build & push a docker image to our hosted container registry (we use quay).
3. Define a my-task-pod.json configuration file.
4. Add the job/task to the scheduler's configuration file.
5. Re-deploy the scheduler to pick up the new configuration

Getting things Running

1. Clone this repo.
2. Run make setup (which uses glide).
3. Set environment variables.
   • CERTS_PATH should be location of your kubernetes credentials.
   • KUBERNETES_SERVICE_HOST should be the host name of your kubernetes cluster.
   • KUBERNETES_SERVICE_PORT should be the port of your kubernetes cluster.
   • SENTRY_DSN should be the dsn to your Sentry project (OPTIONAL).
4. Create a pod definition for your job/task (a sample is here).
5. Create schedule.yml configuration file (a sample is here).
6. Start the service.
   • go run kube-scheduler.go
   • The scheduler takes an optional flag to point to the directory of the schedule.yml configuration file (defaults to current directory).

Gotchas

• The schedule for a task is defined using Go's cron syntax.
• The timezone is local to the scheduler with no way to specify an alternative timezone.
• Active jobs are always unique. The scheduler will prevent a job from starting if the same job is already running.
• An invalid cron syntax will cause a panic in the scheduler from a downstream library
• Duplicate job definitions names will only run the last definition, this is a quirk of yaml parsing
• You can leave a cron as empty string "" to disable an existing job. This does not stop any active configutations/jobs
• The logs produced by the scheduler are any of its own logs and all logs from tasks/jobs.
• We log (and create optional Sentry messages) when a pod fails to start or finish in a "successful" state or when a job is scheduled to start while another instance of the same job is still active.

Copyright and License

Copyright © 2016 Molecule Software, Inc. All Rights Reserved.

Licensed under the MIT License (the "License"). You may not use this work except in compliance with the License. You may obtain a copy of the License in the LICENSE file.