Drain mode is not initiated when terminating a Functionapp in Kubernetes

[yukscck]

Investigative information

When I ran my Functionapp in a Kubernetes cluster, I observed that the Functionapp initiates drain mode and unregisters all the listeners from my application when I delete a Kubernetes pod or scale in my Kubernetes deployment.

However, when I upgraded my Functionapp to use the Azure Functions V4 runtime, I noticed the Functionapp no longer initiates drain mode as before - subsequent messages put into the queues continued to trigger the functions.

Expected behavior

• Functionapp initiates drain mode and unregisters all the listeners from the application when a Kubernetes pod termination occurs.

Actual behavior

• Functionapp continues to execute to handle incoming triggers.

Related information

The above can be reproduced in a personal sample project I have created here:

• Sample codes for containerized functionapp on v4 runtime
• Sample codes for containerized functionapp on v3 runtime
  Note: The only changes are in the Dockerfile and the csproj, as shown in the diffs here

[kshyju]

@liliankasem FYI

[jviau]

Need to verify if this still happens.

[yukscck]

@jviau Thanks for looking into this.

I tried building the container image again today, fresh, and yes, I could still reproduce this issue.

[jviau]

@yukscck thank you for the update. This is on our backlog and we will look into it.