Datadog costs $15/host/month. New Relic is $25/month. For a solo dev on a single VM, that’s overkill. Here’s how we built a health monitoring system that sends daily reports with zero external dependencies.
What We Monitor
Every morning at 7 AM, we get an email report:
=== MisuJob Health Report ===
Database:
Active jobs: 512,847
New today: 3,247
Queue backlog: 142 jobs
Performance:
Avg response time: 89ms
Slow queries (>5s): 2
Error rate: 0.01%
Scheduled Jobs:
Last 24h: 47 executed, 0 failed
Cancelled (PM2 restart): 3
Disk: 14.2 GB / 29 GB (49%)
Memory: 1.1 GB / 2 GB (55%)
The Architecture
class SystemHealthMonitor {
async generateReport(): Promise<HealthReport> {
const [db, perf, jobs, system] = await Promise.all([
this.checkDatabase(),
this.checkPerformance(),
this.checkScheduledJobs(),
this.checkSystemResources(),
]);
return { db, perf, jobs, system };
}
}
Avoiding False Positives
1. Materialized View Refresh Is NOT a Slow Query
async checkSlowQueries(): Promise<SlowQuery[]> {
const result = await pool.query(`
SELECT query, calls, mean_exec_time
FROM pg_stat_statements
WHERE mean_exec_time > 5000 -- > 5 seconds
ORDER BY mean_exec_time DESC
`);
// Filter out known slow-but-expected queries
return result.rows.filter(q =>
!q.query.toLowerCase().includes('refresh materialized view')
);
}
Our MV refresh takes 138 seconds. That’s expected, not an alert.
2. Auto-Cancelled Jobs Are NOT Failures
async checkFailedJobs(): Promise<FailedJob[]> {
const result = await pool.query(`
SELECT job_name, error_message
FROM scheduled_job_executions
WHERE status = 'failed'
AND started_at > NOW() - INTERVAL '24 hours'
`);
// PM2 restarts cancel in-progress jobs - not real failures
return result.rows.filter(j =>
!j.error_message?.includes('Auto-cancelled')
);
}
When PM2 restarts, running scheduled jobs get interrupted. They’re marked “failed” but it’s not a real failure.
Sending the Report
// Run daily at 7 AM UTC
cron.schedule('0 7 * * *', async () => {
const report = await healthMonitor.generateReport();
const html = formatReportAsHtml(report);
await ses.sendEmail({
to: '[email protected]',
subject: `Health Report: ${report.hasIssues ? 'ISSUES' : 'OK'}`,
html
});
});
Total cost: $0 (we already pay for SES).
This monitors MisuJob — 1M+ listings, 50+ importers, all on a single VM.
How do you monitor your side projects? Uptime Robot, custom scripts, or just hope?
