Kong: logging plugins ship a JSON record per request

curl -X POST http://localhost:8001/plugins \
  --data name=http-log \
  --data config.http_endpoint=http://collector.internal:9200/kong \
  --data config.method=POST \
  --data config.timeout=1000 \
  --data config.queue_size=100 \
  --data config.flush_timeout=2

{
  "request":  {"method":"GET","uri":"/billing","size":231},
  "response": {"status":200,"size":1482},
  "latencies":{"request":47,"kong":3,"proxy":44},
  "consumer": {"username":"acme"},
  "service":  {"name":"billing"},
  "started_at": 1766642400123
}

# latencies.kong is Kong's own overhead; latencies.proxy is your upstream.
# That split is the fastest way to answer "is the gateway slow, or is it us".
#
# Batch with queue_size, or a busy gateway makes one HTTP call per request.