Operations

Health

Healthcheck can be found at /health.

Metrics

Prometheus metrics are exposed at /metrics.

ArgoCD connections

These metrics show how the connection to each ArgoCD is holding up:

Metric	What it tells you
argocd_client_retries_total{argocd,method,reason}	Calls to ArgoCD that were repeated after a failure. reason="transient" means the response was lost on the way back, which is usually a reverse proxy or ingress in front of ArgoCD closing the connection. reason="reconnect" means Tangle's own connection had to be re-established.
argocd_client_dials_total{argocd,reason,result}	Connections opened to each ArgoCD, and whether they worked.
argocd_client_connection_generation{argocd}	How many times the connection to each ArgoCD has been replaced since startup.

An occasional transient retry is normal and invisible to users. A steadily rising rate points to the proxy in front of ArgoCD. Check its idle and response timeouts, and whether it's being restarted.