How to Configure Monitoring for Physical and Cloud Servers: From Alerts to On-Call

The value of monitoring isn't in pretty dashboards — it's in warning you before issues hit and helping you pinpoint them after. Set up five core monitoring cate…

Insights:How to Configure Monitoring for Physical and Cloud Servers: From Alerts to On-Call

Five Basic Monitoring Types

Type What to Monitor Common Thresholds
Availability HTTP status codes, response times 2 consecutive failures
Resources CPU, memory, load Exceeding baseline for 5 minutes
Disk Usage rate, I/O 80% / 90% thresholds
Certificate Expiration date 30 days / 7 days
Backup Task success, file sizes Failure or significantly smaller than usual

Set your own baselines instead of copying others.

Alert Channels

  • Email: Good for logging, not suitable for urgent incidents.
  • SMS/Phone: Best for critical outages.
  • Group bots: Ideal for team collaboration.
  • Ticketing: Tracks resolution processes.

Ensure critical incidents can wake people up—don't route alerts to unread inboxes.

On-Call Rules

  • Who receives alerts
  • Required response time
  • Escalation criteria
  • Who has permissions for restarts, rollbacks, traffic switches
  • Post-incident documentation owner

Without clear on-call procedures, monitoring only creates anxiety.

Logs & Tracing

  • Web access log retention period
  • Daily error log rotation
  • Slow query logs
  • Audit trail for critical operations
  • Request IDs linking errors to specific requests

Recovery Drills

Conduct at least quarterly:

  • Restore a database from backup.
  • Switch CDN back to origin once.
  • Roll back code once.
  • Verify certificate renewal.

A recovery process that hasn't been practiced can't be considered reliable.

What Not to Do

  • Don't only monitor the homepage.
  • Don't send all alerts through a single channel.
  • Don't set up monitoring without thresholds or baselines.
  • Don't implement monitoring with no one responsible for responding.
  • Don't substitute monitoring for backups.

Monitoring is the first layer for discovering problems, but recovery capability is the true baseline.

Related Reading: