Skip to main content
Version: 5.x

Monitoring & Integrations

MTE Relay v5 is observed through its structured JSON logs. Every request produces correlated events: performance timings, including the decrypt, validate, proxy, and encrypt phase breakdown; security events such as token failures, replay rejections, and state store problems; an audit record with hashes only, never plaintext; and error events. The events of one request share a correlation_id, which is also the X-MTE-Relay-Request-Id the client sees. A few events from the outbound relay client carry none; the Event Schema Reference lists them.

No vendor SDK runs inside the relay. Your log collector ships the container's output to your platform of choice and adds tags such as the service name, and the prebuilt assets on this site do the rest.

Start here​

  1. Event Schema Reference, the field level contract all integrations build on.
  2. Pick your platform:
PlatformWhat's provided
DatadogLog pipeline, facet and measure list, importable dashboard, five monitors
New RelicImportable two page dashboard, alert policy with five NRQL conditions
GrafanaDashboards for CloudWatch Logs Insights (AWS) and Loki (Kubernetes and on premise)

Splunk, Elastic, and any other JSON capable aggregator work the same way: ingest the JSON lines and build on the schema reference.

  • Any security_event with severity high or critical
  • Any error_event with event_type panic
  • State store health: sustained pair_not_found or any state_restore_failed security events, and relayCode 503 on the "Request completed" line
  • Average performance_event.total_time_ms above your latency budget
  • Spikes in client_pairs_exhausted or invalid_client_id, which usually mean replicas do not share client_id_secret or pair state
  • Relay error codes 470 to 490 and the X-MTE-Relay-Error header at your load balancer; see Errors and recovery

Log volume​

Per request performance events dominate log volume, and the relay emits them for every request. Filter or sample them in your log shipper if you need to; keep security, error, and audit events unsampled.