sentry-observability

Sentry Observability Integration

Overview

Wire Sentry into your logging, metrics, APM, and dashboard toolchain so every error carries full context and every metric correlates back to root-cause events. This skill covers three integration layers: structured logging (winston, pino, structlog) with Sentry event ID correlation, business metrics with error-rate tracking, and cross-tool linking via Sentry Discover, Grafana webhooks, and APM tools.

See also: Logging integration details | Metrics patterns | APM tool cross-linking

Prerequisites

• Sentry SDK v8+ installed (@sentry/node for Node.js, sentry-sdk for Python)
• At least one structured logger configured (winston, pino, or structlog)
• Sentry project DSN available in environment (SENTRY_DSN)
• Dashboard platform accessible (Sentry Discover, Grafana, or Datadog)
• Alert routing strategy decided (who gets paged, where warnings go)

Instructions

Step 1 — Attach Sentry Event IDs to Structured Logs

The core pattern: every log line that triggers a Sentry event carries the event ID, and every Sentry event carries the log context. This creates a two-way link between your log aggregator and Sentry.

Winston (Node.js) — custom transport:

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

class SentryTransport extends winston.Transport {
  log(info: any, callback: () => void) {
    setImmediate(callback);

    if (info.level === 'error' || info.level === 'fatal') {
      const error = info.error instanceof Error
        ? info.error
        : new Error(info.message);

      Sentry.withScope((scope) => {
        scope.setTag('logger', 'winston');
        scope.setContext('log_entry', {
          level: info.level,
          timestamp: info.timestamp,
          service: info.service,
        });
        const eventId = Sentry.captureException(error);
        info.sentry_event_id = eventId;
        info.sentry_url = `https://${process.env.SENTRY_ORG}.sentry.io/issues/?query=${eventId}`;
      });
    }
  }
}

const logger = winston.createLogger({
  defaultMeta: { service: 'api-gateway' },
  transports: [
    new winston.transports.Console({ format: winston.format.json() }),
    new SentryTransport(),
  ],
});

Pino (Node.js) — hooks pattern:

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

const logger = pino({
  hooks: {
    logMethod(inputArgs, method, level) {
      if (level >= 50) { // 50 = error, 60 = fatal
        const [obj, msg] = typeof inputArgs[0] === 'object'
          ? [inputArgs[0], inputArgs[1]]
          : [{}, inputArgs[0]];

        Sentry.withScope((scope) => {
          scope.setTag('logger', 'pino');
          const eventId = Sentry.captureException(
            obj.err instanceof Error ? obj.err : new Error(String(msg))
          );
          if (typeof inputArgs[0] === 'object') {
            inputArgs[0].sentry_event_id = eventId;
          }
        });
      }
      return method.apply(this, inputArgs);
    },
  },
});

For Python structlog integration, see logging-integration.md.

Request ID correlation middleware:

import { randomUUID } from 'crypto';
import * as Sentry from '@sentry/node';

app.use((req, res, next) => {
  const requestId = (req.headers['x-request-id'] as string) || randomUUID();
  req.requestId = requestId;
  res.setHeader('x-request-id', requestId);
  Sentry.setTag('request_id', requestId);
  req.log = logger.child({ requestId, path: req.path });
  next();
});

Step 2 — Correlate Errors with Business Metrics and APM

Connect Sentry events to your metrics pipeline and decide when Sentry performance monitoring is sufficient versus when to add Datadog or New Relic.

Sentry custom metrics (built-in, no extra tools):

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

// Counter — track error rates alongside business events
Sentry.metrics.increment('checkout.attempted', 1, {
  tags: { payment_provider: 'stripe', plan: 'enterprise' },
});

Sentry.metrics.increment('checkout.failed', 1, {
  tags: { payment_provider: 'stripe', failure_reason: 'timeout' },
});

// Distribution — track latency with error correlation
Sentry.metrics.distribution('api.response_time', responseTimeMs, {
  tags: { endpoint: '/api/orders', status_code: String(res.statusCode) },
  unit: 'millisecond',
});

// Gauge — track queue depth, connection pool size
Sentry.metrics.gauge('db.pool.active', pool.activeCount, {
  tags: { database: 'primary' },
});

// Set — track unique affected users during incidents
Sentry.metrics.set('incident.affected_users', userId, {
  tags: { incident: 'payment-outage-2026-03' },
});

For Prometheus dual-write patterns, see metrics-integration.md.

When to use Sentry performance vs Datadog/New Relic:

Scenario	Use Sentry Performance	Use Datadog/New Relic
Frontend + backend in one view	Yes — unified error + perf traces	Overkill if Sentry covers your stack
Infrastructure metrics (CPU, memory)	No — Sentry does not collect infra	Yes — native host agent collection
100+ custom metric series	Limited query constraints	Yes — built for high-cardinality
Budget-constrained, < 5 services	Yes — one tool, one bill	Unnecessary cost

Datadog + Sentry cross-linking via beforeSend:

import tracer from 'dd-trace';
import * as Sentry from '@sentry/node';

// dd-trace MUST be initialized before @sentry/node
Sentry.init({
  dsn: process.env.SENTRY_DSN,
  beforeSend(event) {
    const span = tracer.scope().active();
    if (span) {
      const traceId = span.context().toTraceId();
      event.tags = { ...event.tags, 'dd.trace_id': traceId };
      event.contexts = {
        ...event.contexts,
        datadog: {
          trace_url: `https://app.datadoghq.com/apm/trace/${traceId}`,
          trace_id: traceId,
        },
      };
    }
    return event;
  },
});

For New Relic correlation patterns, see apm-tool-integration.md.

Step 3 — Build Dashboards and Connect External Tools

Use Sentry Discover for error analytics, set up Grafana webhooks for unified dashboards, and link Sentry events to external tools via setContext.

Linking all tools via Sentry.setContext('monitoring', ...):

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

function setMonitoringContext(req: Request) {
  const traceId = Sentry.getActiveSpan()?.spanContext().traceId;
  const spanId = Sentry.getActiveSpan()?.spanContext().spanId;
  const requestId = req.headers['x-request-id'] as string || crypto.randomUUID();

  // setContext creates a named section in the Sentry event sidebar
  Sentry.setContext('monitoring', {
    traceId,
    spanId,
    requestId,
    grafana_dashboard: `https://grafana.example.com/d/abc123?var-trace_id=${traceId}`,
    kibana_logs: `https://kibana.example.com/app/logs?query=request_id:${requestId}`,
    datadog_trace: traceId
      ? `https://app.datadoghq.com/apm/trace/${traceId}`
      : undefined,
  });

  Sentry.setTag('request_id', requestId);
  Sentry.setTag('trace_id', traceId || 'none');
  Sentry.setTag('deployment', process.env.DEPLOYMENT_ID || 'unknown');
}

app.use((req, res, next) => {
  setMonitoringContext(req);
  next();
});

Grafana integration via Sentry webhooks:

Configure in Settings > Integrations > Internal Integrations. Point the webhook URL at a receiver that transforms Sentry events into Grafana annotations:

// Receive Sentry webhook, create Grafana annotation
app.post('/sentry-to-grafana', async (req, res) => {
  const { event } = req.body;
  if (!event) return res.status(200).send('ignored');

  await fetch(`${process.env.GRAFANA_URL}/api/annotations`, {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      Authorization: `Bearer ${process.env.GRAFANA_API_KEY}`,
    },
    body: JSON.stringify({
      dashboardUID: process.env.GRAFANA_DASHBOARD_UID,
      panelId: 1,
      time: new Date(event.datetime).getTime(),
      tags: ['sentry', event.level, event.project],
      text: `**${event.title}**\nLevel: ${event.level}\n[View in Sentry](${event.web_url})`,
    }),
  });

  res.status(201).json({ status: 'annotation_created' });
});

Alert routing across tools:

Issue Alert: "Critical Production Error"
  When: An event is first seen
  If: level is fatal AND environment is production
  Then: PagerDuty (Critical) + #alerts-critical Slack + Grafana annotation
  Frequency: Once per issue

Metric Alert: "Error Rate Spike"
  When: Error count > 50 in 5 minutes
  Then: PagerDuty (High) + #alerts-production Slack + webhook to Grafana
  Resolve: Error count < 5 for 10 minutes

Metric Alert: "Latency Regression"
  When: p95(transaction.duration) for /api/* > 2000ms for 10 minutes
  Then: #alerts-performance Slack + JIRA ticket via webhook
  Resolve: p95 < 1000ms for 15 minutes

Output

After completing these steps you will have:

• Winston/pino/structlog forwarding errors to Sentry with event IDs stamped into log lines
• Sentry custom metrics (counters, gauges, distributions, sets) tracking business KPIs
• beforeSend hooks linking Sentry events to Datadog traces and New Relic transactions
• Sentry.setContext('monitoring', { traceId, spanId }) linking every event to external tool URLs
• Grafana annotations created from Sentry webhooks on infrastructure dashboards
• Tiered alert routing: fatal errors page on-call, warnings go to Slack, latency issues create tickets

Error Handling

Error	Cause	Solution
Sentry event IDs missing from logs	Transport/processor not wired up	Verify SentryTransport is in Winston transports o

---

Content truncated.