Salesforce observability

Salesforce Observability

Overview

Instrument Salesforce integrations with API limit monitoring, SOQL performance tracking, error classification, and alerting. Uses Salesforce's built-in Limits API and EventLogFile for deep visibility.

Prerequisites

• jsforce connection configured
• Prometheus or compatible metrics backend (optional)
• Grafana or similar dashboarding tool (optional)
• Salesforce Enterprise+ for EventLogFile access

Instructions

Step 1: API Limit Monitoring (Core Metric)

import { getConnection } from './salesforce/connection';
import { Registry, Gauge, Counter, Histogram } from 'prom-client';

const registry = new Registry();

// The single most important Salesforce metric
const apiLimitGauge = new Gauge({
  name: 'salesforce_api_limit_remaining',
  help: 'Remaining daily API calls',
  registers: [registry],
});

const apiLimitMaxGauge = new Gauge({
  name: 'salesforce_api_limit_max',
  help: 'Maximum daily API calls',
  registers: [registry],
});

const apiUsagePercent = new Gauge({
  name: 'salesforce_api_usage_percent',
  help: 'Percentage of daily API calls used',
  registers: [registry],
});

// Poll limits every 5 minutes (each poll = 1 API call)
setInterval(async () => {
  try {
    const conn = await getConnection();
    const limits = await conn.request('/services/data/v59.0/limits/');

    apiLimitGauge.set(limits.DailyApiRequests.Remaining);
    apiLimitMaxGauge.set(limits.DailyApiRequests.Max);

    const used = limits.DailyApiRequests.Max - limits.DailyApiRequests.Remaining;
    apiUsagePercent.set((used / limits.DailyApiRequests.Max) * 100);
  } catch (error) {
    console.error('Failed to poll SF limits:', error);
  }
}, 5 * 60 * 1000);

Step 2: Request Instrumentation

const sfRequestDuration = new Histogram({
  name: 'salesforce_request_duration_seconds',
  help: 'Salesforce API request duration',
  labelNames: ['operation', 'sobject'],
  buckets: [0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10],
  registers: [registry],
});

const sfRequestCounter = new Counter({
  name: 'salesforce_requests_total',
  help: 'Total Salesforce API requests',
  labelNames: ['operation', 'sobject', 'status'],
  registers: [registry],
});

const sfErrorCounter = new Counter({
  name: 'salesforce_errors_total',
  help: 'Salesforce errors by error code',
  labelNames: ['error_code', 'sobject'],
  registers: [registry],
});

// Instrumented wrapper for all SF operations
async function instrumentedSfCall<T>(
  operation: string,
  sobject: string,
  fn: () => Promise<T>
): Promise<T> {
  const timer = sfRequestDuration.startTimer({ operation, sobject });

  try {
    const result = await fn();
    sfRequestCounter.inc({ operation, sobject, status: 'success' });
    return result;
  } catch (error: any) {
    const errorCode = error.errorCode || 'UNKNOWN';
    sfRequestCounter.inc({ operation, sobject, status: 'error' });
    sfErrorCounter.inc({ error_code: errorCode, sobject });
    throw error;
  } finally {
    timer();
  }
}

// Usage
const accounts = await instrumentedSfCall('query', 'Account', () =>
  conn.query('SELECT Id, Name FROM Account LIMIT 10')
);

Step 3: Salesforce-Native Monitoring (EventLogFile)

// EventLogFile provides detailed API usage data (Enterprise+ only)
// Available event types: API, Login, Logout, URI, BulkApi, etc.

async function getApiUsageEvents(days: number = 1) {
  const conn = await getConnection();

  const events = await conn.query(`
    SELECT Id, EventType, LogDate, LogFileLength
    FROM EventLogFile
    WHERE EventType IN ('API', 'RestApi', 'BulkApi')
      AND LogDate >= LAST_N_DAYS:${days}
    ORDER BY LogDate DESC
  `);

  for (const event of events.records) {
    // Download and parse the CSV log file
    const logContent = await conn.request(
      `/services/data/v59.0/sobjects/EventLogFile/${event.Id}/LogFile`
    );
    // Parse CSV to extract: USER_ID, URI, METHOD, STATUS_CODE, RUN_TIME, CPU_TIME
    console.log(`${event.EventType} on ${event.LogDate}: ${event.LogFileLength} bytes`);
  }
}

Step 4: Structured Logging

import pino from 'pino';

const logger = pino({ name: 'salesforce-integration' });

function logSfOperation(
  operation: string,
  sobject: string,
  details: Record<string, any>,
  durationMs: number
) {
  logger.info({
    service: 'salesforce',
    operation,
    sobject,
    durationMs,
    ...details,
    // Parse Sforce-Limit-Info header from response
    // Format: "api-usage=135/150000"
  });
}

Step 5: Alert Rules

# prometheus-alerts.yaml
groups:
  - name: salesforce_alerts
    rules:
      - alert: SalesforceApiLimitCritical
        expr: salesforce_api_usage_percent > 90
        for: 5m
        labels:
          severity: critical
        annotations:
          summary: "Salesforce API usage above 90% ({{ $value }}%)"
          description: "API calls will be blocked at 100%. Reduce usage or contact Salesforce for limit increase."

      - alert: SalesforceApiLimitWarning
        expr: salesforce_api_usage_percent > 75
        for: 15m
        labels:
          severity: warning
        annotations:
          summary: "Salesforce API usage above 75%"

      - alert: SalesforceHighErrorRate
        expr: |
          rate(salesforce_errors_total[5m]) /
          rate(salesforce_requests_total[5m]) > 0.05
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "Salesforce error rate > 5%"

      - alert: SalesforceHighLatency
        expr: |
          histogram_quantile(0.95,
            rate(salesforce_request_duration_seconds_bucket[5m])
          ) > 5
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "Salesforce P95 latency > 5s"

      - alert: SalesforceAuthFailure
        expr: increase(salesforce_errors_total{error_code="INVALID_SESSION_ID"}[5m]) > 0
        labels:
          severity: critical
        annotations:
          summary: "Salesforce authentication failures detected"

Key Salesforce Metrics to Monitor

Metric	Source	Alert Threshold
Daily API remaining	/services/data/v59.0/limits/	< 10% remaining
Request latency P95	Instrumented client	> 5 seconds
Error rate by code	Instrumented client	> 5%
Bulk API job failures	Bulk job results	Any failures
Session/token expiry	Auth error count	Any INVALID_SESSION_ID
Data storage used	Limits API	> 90% capacity

Error Handling

Issue	Cause	Solution
Limits poll failing	Token expired	Auto-refresh connection
High cardinality	Too many label values	Use error_code, not error message
Missing EventLogFile	Not Enterprise+	Use instrumented client instead
Alert storms	Threshold too low	Tune thresholds with historical data

Resources

• Limits REST Resource
• EventLogFile
• Salesforce Status API
• Prometheus Best Practices

Next Steps

For incident response, see salesforce-incident-runbook.