Self-hosted monitoring step by step
A green container does not prove that users can reach a working service. Combine external checks, host metrics, logs and backup recovery tests.
Monitoring should answer the user's question: does the service work correctly from outside? Container state and CPU use matter, but cannot reveal bad DNS, an expired certificate or a broken login path.
Start with a small number of alerts that have clear owners and responses. Excess notifications quickly teach people to ignore important ones.
External checks
Monitor the public URL, expected status and optionally response content. Check DNS, certificates, critical ports and email dependencies separately.
- Match frequency to the real recovery objective.
- Notify on failure and recovery.
- Confirm with multiple attempts to reduce noise.
Host, containers and data
Collect disk, memory, CPU, restart state and application errors. For databases, observe connections, query time and backup outcome.
Incident and escalation
Every alert should name the target, time, last healthy result and a simple response. Group repeats and retain state-change history.
- Never include secrets in alerts.
- Separate warning from outage.
- After an incident, improve the threshold or runbook as well as the service.
Minimum monitoring
- The URL is checked off-host.
- Certificate and DNS have separate alerts.
- Disk, container restarts and backup are observed.
- Every alert has an owner and runbook.
- Incident history reveals trends.
Put it into practice
Build a domain profile
Combine seven checks into one report and enable daily technical alerts.
Build a domain profileFrequently asked questions
Does one configuration fit every environment?
No. Inventory traffic, dependencies and the threat model first, then tighten controls. Verify each change externally and keep a rollback plan.
Where should I start?
Start with a tested backup, current patches and a smaller public attack surface. Then add monitoring and document an owner for every service.
How often should I repeat the review?
After every infrastructure or major-version change and at least quarterly. Monitor critical dependencies, certificates and backup outcomes continuously.