Jafra Analyzer

The Jafra Analyzer is the central service for JFR recording data. It receives completed chunks from every Agent, verifies their integrity, stores durable chunk payloads, stitches recordings on demand for analysis, applies JDK Mission Control rules, and exposes JVM evidence through HTTP APIs.

It is a Java 21 service built with Quarkus and JDK Mission Control libraries.

What the Analyzer does

For every upload, the Analyzer:

1. validates workload metadata and protocol version;
2. enforces ordered frame offsets;
3. verifies the final SHA-256 checksum;
4. commits the chunk payload and .meta identity to persistent storage;
5. acknowledges the Agent only after that identity is durable.

This commit-before-acknowledgement behavior allows Agents to retry without creating duplicate data. Contiguous reconstructions for HTTP analysis are built later into a bounded stitch cache, not as a second durable copy on ingest.

Analysis experiences

• Recording inventory shows available files by namespace, Pod, and container.
• Automated reports run JDK Mission Control rules and return scored findings.
• Event summaries expose counts and statistics from the event types present in the recording.
• Time-window queries combine the rotations that overlap an incident period, with stitch-cache reuse and event filtering for sliding windows.

See Use the Analyzer API for request examples.

Configuration

Property	v0.0.2 value	Purpose
quarkus.http.port	8080	HTTP API
quarkus.grpc.server.port	9090	gRPC ingest
quarkus.http.host	0.0.0.0	HTTP bind address
quarkus.grpc.server.host	0.0.0.0	gRPC bind address
jafra.ingest.max-metadata-length	4096	Ingest metadata limit
jafra.ingest.max-active-streams	32	Concurrent ingest streams
jafra.storage.root	/var/lib/jafra/analyzer	Persistent storage root
jafra.storage.stitch-cache.max-bytes	2147483648 (2 Gi)	Stitch-cache budget
jafra.storage.stitch-cache.ttl	PT2M	Unused entry TTL

The supplied Deployment runs one replica, mounts a 5 GiB ReadWriteOnce PVC, sets heap options -Xms256m -Xmx1024m, and limits container memory to 2 GiB. On startup, it removes incomplete transfers, reloads durable identities, and clears legacy durable stitched.jfr files if present.

HTTP endpoints

• GET /health
• GET /q/health
• GET /q/metrics
• GET /api/v1/status
• GET /api/v1/recordings
• GET /api/v1/namespaces/{namespace}/pods/{pod}/containers/{container}
• GET .../report
• GET .../summary
• GET .../recordings/{filename}/report
• GET .../recordings/{filename}/summary

Report endpoints return JMC rule results. Summary endpoints return raw event aggregates. Both support recording selection and time-window filtering.

Metrics and logs

Prometheus metrics are available through /q/metrics, including accepted, duplicate, rejected, checksum-failure, and protocol-failure ingest counters.

Chunk and stitch messages include JSON metadata strings in log lines.