# Found the root cause. Turns out my disk speed is slow enough that it takes longer to start the server than the fixed start_period and interval value. Basically health check fires too early. Fixed it by increasing start_period in docker compose.