A Packer-built Compute Engine image was created by CI but won't boot. How should you investigate the failure with the CI/CD team?

The most efficient way to investigate a non-booting Compute Engine VM is to examine its serial port output. This output often contains critical boot-time messages from the operating system, including errors, kernel panics, or issues with bootloaders. Google Cloud automatically streams serial port output to Cloud Logging, making it accessible even if the VM fails to fully boot. Verifying Cloud Logging access ensures you can retrieve these crucial logs. Creating a new feature branch and rebuilding the image is premature; you need to diagnose the root cause first. Shutting down and exporting the disk is a more complex and time-consuming process that might be necessary for deeper forensics but isn't the first step for a boot failure. Installing Packer locally and rebuilding the image in a personal project is also not the most direct approach; the issue might be specific to the CI environment or the image's interaction with Compute Engine.