Production app on Compute Engine managed instance group behind an HTTP(S) load balancer is inaccessible. Which monitoring technique should alert you when the application is unavailable?

Stackdriver (now Cloud Monitoring) uptime checks are the correct choice because they simulate external user traffic to your application's public endpoint, directly verifying its availability from outside your infrastructure. If the application is inaccessible, these checks will fail and trigger an alert. Cloud Load Balancing health checks and Managed instance group health checks primarily monitor the health of the instances themselves, not necessarily the application running on them. An instance could be healthy (running) but the application might still be down or misconfigured. Smoke tests are typically run as part of a deployment pipeline or manually, not as continuous, external availability monitoring.