feat(node): Add `nodeRuntimeMetricsIntegration`

[github-actions]

Note

The pull request "feat(node): Add nodeRuntimeMetricsIntegration" was created by @chargome but did not reference an issue. Therefore this issue was created for better visibility in external tools like Linear.

Adds nodeRuntimeMetricsIntegration to @sentry/node and @sentry/node-core. When enabled, the integration periodically collects Node.js runtime health metrics and emits them to Sentry via the metrics pipeline.

Usage

import * as Sentry from '@sentry/node';

Sentry.init({
 dsn: '...',
 integrations: [
 Sentry.nodeRuntimeMetricsIntegration(),
 ],
});

Default metrics (8)

Emitted every 30 seconds out of the box:

Metric	Type	Unit	Description
node.runtime.mem.rss	gauge	byte	Resident Set Size — actual process memory footprint
node.runtime.mem.heap_used	gauge	byte	V8 heap currently in use — tracks GC pressure and leaks
node.runtime.mem.heap_total	gauge	byte	Total V8 heap allocated — paired with heap_used to see headroom
node.runtime.cpu.utilization	gauge	—	CPU time / wall-clock time ratio (can exceed 1.0 on multi-core)
node.runtime.event_loop.delay.p50	gauge	second	Median event loop delay — baseline latency
node.runtime.event_loop.delay.p99	gauge	second	99th percentile event loop delay — tail latency / spikes
node.runtime.event_loop.utilization	gauge	—	Fraction of time the event loop was active
node.runtime.process.uptime	counter	second	Cumulative uptime — useful for detecting restarts / crashes

Opt-in metrics (off by default)

Sentry.nodeRuntimeMetricsIntegration({
 collect: {
 cpuTime: true, // node.runtime.cpu.user + node.runtime.cpu.system (raw seconds)
 memExternal: true, // node.runtime.mem.external + node.runtime.mem.array_buffers
 eventLoopDelayMin: true,
 eventLoopDelayMax: true,
 eventLoopDelayMean: true,
 eventLoopDelayP90: true,
 },
})

Any default metric can also be turned off:

Sentry.nodeRuntimeMetricsIntegration({
 collect: {
 uptime: false,
 eventLoopDelayP50: false,
 },
})

Collection interval

Sentry.nodeRuntimeMetricsIntegration({
 collectionIntervalMs: 60_000, // default: 30_000
})

Serverless (Next.js on Vercel, AWS Lambda, etc.)

In serverless environments the process is short-lived, so periodic intervals may never fire. The integration handles this automatically:

• A beforeExit listener fires at the end of every invocation, collects a final snapshot of all metrics, and calls flushIfServerless() — which is a no-op in long-running servers but triggers an immediate flush on serverless platforms.
• The setInterval is unref()-ed so it never prevents the process from exiting between invocations.
• No special configuration needed — the same nodeRuntimeMetricsIntegration() call works in both environments.

Runtime compatibility

This integration is Node.js only. Bun and Deno will be addressed in separate integrations that use their respective native APIs.

[linear-code]

JS-1997