high availability: deploy Repairnator in Kubernetes

Core design:

• A queue system like ActiveMQ to save build ids
• A K8s cron job running repairnator scaner to get latest build ids and add them into the queue
• A set of pods running repairnator pipeline which fetches build id from the queue and does the analysis

Such an architecture could decrease coupling and take advantages of Kubernetes features. Since Repairnator has already implemented different Docker images, it is not very difficult to push forward.

Tasks (one per pull requests):

• Rename BuildRunner into DockerPipelineRunner
• Create empty ActiveMQPipelineRunner extends AbstractPoolManager (branch by abstraction strategy)
• Change field name of DockerPipelineRunner to pipelineRunner
• Add interface PipelineRunner for ActiveMQPipelineRunner to implements
• Implement ActiveMQPipelineRunner
• Add NopeRunner which does nothing. It would allow to run the scanner only, even without docker installed.
• NopeRunner should be the default
• Add support for configuring the runner to use (ie either DockerPipelineRunner or ActiveMQPipelineRunner)
• Change Travis CI config to deploy an ActiveMQ there and test the core functionality

[gluckzhang]

Very good. Instead of adding things here, what about breaking this PR in smaller ones, to be merged one after the other? This is the “branch by abstraction” practice.

Hi @monperrus , good idea. I also find that the PR is growing too fat now. We will break it into small ones 😃

[monperrus]

Very good. Instead of adding things here, what about breaking this PR in smaller ones, to be merged one after the other? This is the “branch by abstraction” practice.