Insights:Incident Response and Troubleshooting: What to Check First — A 5-Step Method
Five-Step Sequence
| Step | Check | Reason |
|---|---|---|
| 1 | Impact Scope | Determines Severity |
| 2 | Recent Changes | Most Common Trigger |
| 3 | Resources | Disk / Memory / Connections |
| 4 | Dependencies | Database / Third-party Services |
| 5 | Logs | Pinpoint Specific Errors |
I. Determine Impact Scope
- Entire site or only certain pages.
- All users or specific regions.
- Whether alternative paths exist.
The impact scope decides whether someone needs to be woken up immediately.
II. Check Recent Changes
- Has anything been deployed in the last 24 hours?
- Were any configurations or permissions modified?
- Were any dependencies updated?
If there have been changes, consider rolling back first.
III. Check Resources
- Is the disk full?
- Is memory exhausted?
- Are connections maxed out?
- Is CPU usage abnormal?
Resource problems usually show up as "sudden slowdown" or "intermittent failures."
IV. Check Dependencies
- Is the database reachable?
- Is the cache service healthy?
- Are third-party API calls timing out?
Dependency failures are often misdiagnosed as "our own code problem."
V. Check Logs
- Recent error log entries.
- Anomalies at specific timestamps.
- Recurring error patterns.
At this step, you should typically be able to pinpoint the specific code or configuration.
Three Rules During Troubleshooting
- Stop the bleeding first, then dig deeper.
- Record conclusions at every step to avoid repeating排查.
- Escalate when uncertain—don't tough it out.
Scope
This article is organized around common web architectures. Distributed or multi-service architectures require distributed tracing and follow a different排查 path.
Related Reading: