Intermittent 502 errors on Autopilot

[maxwbwu]

We currently use Autopilot and there's no events fired for endpoints, services, pods objects etc. Nothing failed health checks. But the loadbalancer logged errors of failed_to_pick_backend. Is there anyway debug this further? I want to see endpoint group logs/events on gcp but i don't see anyway to.

[Marvin_Lucero · Accepted solution]

Hi @maxwbwu ,

Based from the error "failed_to_pick_backend" in the logs, this means that the load balancer failed to select what backend will be used to route for the incoming traffic. If you want to debug this error, you can consider to check the following :

1. Backend Service Configuration

a. Verify if the appropriate backend (Compute engine or Cloud Run) to your backend service configuration.
b. Check if the backend service is listening on the correct port. Make sure that it matches the LB's forwarding rule.

2. You may also check the firewall rules and ensure that backend service's instances are allowing traffic from the LB. 

You may also consult support and create a ticket for it. Your project and configurations might need a deeper investigation to pin-point what's causing the "failed_to_pick_backend".