What is Stop-the-World in GC?

Direct Answer

A stop-the-world pause is an interval when Java application, or mutator, threads in one JVM are suspended while collector or other JVM work continues. In-process requests that overlap the pause cannot execute Java code during that interval, so the pause can affect latency.

Code Example

@RestController
public class OrderController {
    private final OrderService orderService;

    @GetMapping("/orders/{id}")
    public ResponseEntity<Order> getOrder(@PathVariable Long id) {
        long start = System.nanoTime();
        var order = orderService.findById(id);
        long elapsedNanos = System.nanoTime() - start;
        // elapsedNanos includes any STW interval that overlapped this work.
        return ResponseEntity.ok(order);
    }
}

Use a monotonic clock for elapsed time and correlate request telemetry with unified GC logs, for example -Xlog:gc*:file=gc.log:time,uptime. One slow request does not prove that GC caused it.

Common Mistakes & Pitfalls

• Calling all GC work background work: Collectors can perform concurrent work, but they also use stop-the-world phases.
• Blaming every latency spike on GC: Correlate request traces, CPU, safepoint/GC logs, and allocation data before deciding on a cause.
• Promising a collector-specific pause number: Heap, live set, allocation, CPU, JVM version, flags, and workload all matter; benchmark the actual service.