Question

You manage a managed online endpoint that currently serves 100% of production traffic from a deployment named 'blue'. After deploying a new version named 'green' and gradually shifting 50% of traffic to it, monitoring shows 'green' is returning elevated error rates and increased latency. You need to safely roll back so that all traffic returns to the known-good deployment as quickly as possible while preserving the ability to investigate 'green'. What should you do?

Accepted Answer

Update the endpoint traffic allocation to route 100% to 'blue' and 0% to 'green'.

Answer

Delete the 'green' deployment immediately, which automatically routes all traffic back to 'blue'.

Answer

Create a new deployment from the 'blue' image and split traffic evenly between it and 'green'.

Answer

Scale the 'green' deployment instance count to zero and leave the traffic allocation unchanged.