Maintenance Windows: Automatic Alert Silence

Every deployment, every server update, every database maintenance potentially generates alerts. These alerts are expected - your service is temporarily unavailable because you planned it that way. Yet, without proper configuration, these predictable alerts pollute your flow and create alert fatigue.

Maintenance windows solve this problem. They allow automatically suspending alerts during a defined period, avoiding unnecessary noise while continuing to collect monitoring data for your statistics.

This guide shows you how to effectively configure maintenance windows in MoniTao. From one-time maintenance to recurring deployments, you'll learn to properly manage these planned interruptions.

What is a Maintenance Window?

A maintenance window is a period during which monitoring behaves differently:

• Suspended alerts: No notifications are sent during the window, even if the service is detected as unavailable. Your team isn't disturbed for planned downtime.
• Active monitoring: Checks continue to run and data is recorded. You keep visibility on actual unavailability duration.
• Preserved statistics: Availability reports can include or exclude planned maintenance according to your configuration. Distinguish planned downtime from unplanned.
• Automatic resume: At the end of the window, alerts automatically resume. If the service hasn't returned, you're immediately notified.

When to Use a Maintenance Window?

Maintenance windows are useful in many scenarios:

• Recurring deployments: Your team deploys every Tuesday at 10am? Configure an automatic weekly window from 10am to 11am every Tuesday.
• Server maintenance: OS update, reboot to apply patches. Create a one-time window covering the estimated duration.
• Database migration: Major migrations can require several hours. A maintenance window avoids dozens of alerts during the operation.
• Provider maintenance: Your host announces network maintenance Saturday 2am-4am. Schedule a corresponding window for all affected monitors.
• Load testing: Performance tests can slow the service or make it temporarily unavailable. A window avoids false latency alerts.

Configure a Maintenance Window

MoniTao allows creating maintenance windows in several ways:

1. One-time maintenance: From the dashboard, click "Maintenance" then "New window". Define start/end date/time, select affected monitors.
2. Recurring maintenance: Check "Recurring" and define the pattern (daily, weekly, monthly). For example: "Every Tuesday from 10am to 11am".
3. Maintenance by tag: Apply maintenance to all monitors with a tag (e.g., "production-web"). New monitors with this tag are automatically included.
4. Maintenance via API: Integrate maintenance creation in your deployment scripts. The MoniTao API allows creating a window just before deployment.

Configuration Examples

Here are common maintenance window examples:

# Recurring maintenance for weekly deployment
maintenance:
  name: "Tuesday deployment"
  schedule: "0 10 * * 2"  # Tuesday 10am (cron)
  duration: 60  # minutes
  monitors: ["tag:production"]

# One-time maintenance for DB migration
maintenance:
  name: "PostgreSQL 15 Migration"
  start: "2024-02-15 02:00:00"
  end: "2024-02-15 06:00:00"
  monitors: ["db-primary", "db-replica", "api-prod"]

# Provider maintenance (notification received)
maintenance:
  name: "OVH Network Maintenance"
  start: "2024-02-17 03:00:00"
  end: "2024-02-17 05:00:00"
  monitors: ["tag:ovh-servers"]

These configurations show different use cases: cron recurrence for regular deployments, fixed window for one-time operations, and tag application for maintenance affecting a server group.

Types of Maintenance Windows

MoniTao supports several window types based on your needs:

• Complete silence: No alerts are sent regardless of severity. Default option for planned maintenance.
• Critical alerts only: Only the most severe alerts (e.g., complete timeout) are sent. Useful to detect if maintenance exceeds plan.
• Suspended monitoring: Checks are completely stopped. Useful if the checks themselves interfere with maintenance.
• Statistics exclusion: Downtime during the window isn't counted in availability statistics. Your SLAs aren't impacted by planned maintenance.

Best Practices

Optimize maintenance window usage:

• Safety margin: Plan 20-30% margin over estimated duration. A 30-minute deployment deserves a 45-minute window.
• Post-maintenance check: Configure an immediate check at window end to quickly detect if service hasn't properly returned.
• Team communication: Notify the team when a maintenance window starts. MoniTao can send a summary of monitors in maintenance.
• Regular audit: Review recurring windows monthly. Deployments no longer using their usual slot create unnecessary windows.

Frequently Asked Questions

Conclusion

Maintenance windows are essential for healthy monitoring. They eliminate predictable noise, reduce alert fatigue, and let your team focus on real incidents.

MoniTao gives you all the tools to manage your maintenance: one-time or recurring windows, tag application, API integration for automation. Configure your first window and enjoy serene deployments.

Ready to Sleep Soundly?

Start free, no credit card required.

Create My Account See pricing