Initial implementation for a graceful shutdown of the watchdog.

[Templum]

As discussed in #573 currently the watchdog does not react upon receiving a SIGTERM. I used the new functionality of the HTTP server to perform a graceful shutdown once a SIGTERM was received. As windows container are more rarely it currently only supports Linux container.

Description

I created a new function named gracefulShutdown which expects a pointer to an http server which will be used to perform the shutdown invocation. Further for testing purposes it also accepts a bool channel. However, I'm not quite happy with that trick, so if someone knows a better approach I'm happy to learn.

Motivation and Context

• I have raised an issue to propose this change Architecture: watchdog - graceful termination #573 & During downscaling the pods don't get killed gracefully faas-netes#156

How Has This Been Tested?

I created a simple test case which uses an HTTP Server to check if a received SIGTERM leads to a shutdown invocation on the HTTP Server.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I've read the CONTRIBUTION guide
• I have signed-off my commits with git commit -s
• I have added tests to cover my changes.
• All new and existing tests passed.

[alexellis]

Hi Simon, in the referenced issue (#573) I asked for us to use the idomatic Golang approach of cancel with the HTTP server implementation.

I've raised a PR that makes use of the pattern from the Golang documentation and have tested it inside Docker Swarm and Kubernetes. Please test it and report back if you have any feedback.

[Templum]

Hi Alex,

as you pointed towards the usage of the HTTP server, I also implemented it. Overall our implementations are kind if similar with some minor differences, for example, you used a flag and used it within the health check. I, on the other hand, used another sort of context avoiding the extra timer. Can you describe me how you tested it? So I can test it in the same way.