Rancher 1.6.x service upgrade issues

1

Is anyone else having problems with upgrading services in their stacks on Rancher 1.6?

I often get errors like this:
“Failed to allocate instance [container:1i113198]: Bad instance [container:1i113198] in state [error]: Allocation failed: Host must satisfy resource constraints: [60545:3000/tcp portReservation]. needs host with label environment.type: production”

This is trying to do a simple upgrade of a service in a stack, where the service stops and then pulls a new image and spins up a new container.

Often, rolling back then also fails, and you have to delete the service in the hope that is gets re-created again by Rancher.

This seems to have been a problem since at least 1.6.15.

2

No-one else seeing this problem?

Looks like there is a real problem with freeing up the ports that containers have been using before containers in a stack are upgraded - even efter they have been stopped. When trying to start the upgraded container, Rancher has is trying to use the port again (e.g. 54051:3000/tcp) - which is failing, and which makes the whole upgrade process fail.