v4.6.2

Rust-Java REST 4.6.2

English | Turkish

4.6.2 fixes the aggregate interval that could be lost when Glowroot Central closed a retained
HTTP/2 connection at its 20-minute maximum age. REST annotations, handlers, services, records,
validation, and business logic are unchanged.

Maven Dependency

<dependency>
  <groupId>com.reactor</groupId>
  <artifactId>rust-java-rest</artifactId>
  <version>4.6.2</version>
</dependency>

Reliable Collector Rotation

Stage	Behavior
Before snapshot	The standalone exporter rotates at 15 minutes and checks connection readiness before draining counters.
Close race	Rust reconnects, sends collector init, and retries the same protobuf bytes.
Before ACK	One bounded aggregate batch remains owned by Rust.
Longer outage	New transactions remain in preallocated counters; no unbounded interval queue is created.
Recovery	Central ACK releases the batch and the pending transaction gauge returns to zero.

For embedded Rust-Java REST telemetry, the exporter still opens a bounded connection only for the
export window. The ACK-safe retry protects transient closes without adding a retained idle
connection beside Hyper.

Operational Check

Inspect /diagnostics/glowroot or /metrics:

• collector_rotations should increase near 15 minutes in standalone mode.
• collector_request_retries can increase during a close race.
• pending_aggregate_transactions must return to zero after recovery.
• Recovered rotations must not increase dropped_transactions, dropped_intervals, or
  last_error_code.

The retry state is bounded and in memory. Process termination before Central acknowledgement is not
durable; a disk WAL would be a separate feature with a different memory and I/O contract.

Compatibility

Component	Version
Rust-Java REST	4.6.2
REST ABI	29
Glowroot ABI	6
Spring agent	java-rust-glowroot-agent:0.5.3
Linux x64	glibc 2.17 or newer

The native DLL/SO is built from clean rust-spring revision
5bc1b825673967b8105c05944523f977131071de. Use the binary packaged by this Maven version.