The Prometheus backend behind api.octestbed.org has been returning 502 Bad Gateway on all authenticated /api/v1/* endpoints since at least 2026-04-22T06:47Z UTC. Confirmed from both inside the cluster (pc179) and externally (home network).
Diagnostics:
- DNS, TLS, network: all OK
- nginx auth layer: working (401 without key, 403 with bad key, 502 with valid key)
- ALL endpoints fail: /query, /query_range, /labels, /status/config, /targets
- ALL metrics fail: up, outletPower, outletVoltage, outletCurrent, inletPower
- Port 9090 (Prometheus default): CLOSED/FILTERED
- Port 443: OPEN, Port 80: OPEN
The nginx reverse proxy is healthy but cannot reach the upstream Prometheus service. Could you please check if the Prometheus process needs to be restarted?
This is blocking power measurement experiments on the FPGA cluster. Thank you.