We have a setup of the coder control plane together with a metric collector sidecar in ECS (2 containers in the same task)
metrics are exposed to the local port. (2112) and the collector listens to http://localhost:2112/metrics to scrape the metrics and send them to our metrics repository (datadog in our case)
when I enabled Prometheus metrics using CODER_PROMETHEUS_ENABLE everything worked just fine and all the metrics arrived to our metrics repository.
but when I enabled the option to get agent stats metrics using CODER_PROMETHEUS_COLLECT_AGENT_STATS we start getting 500 from http://localhost:2112/metrics

Is there an additional configuration that needs to be changed to make that work?