Hands-on AWS knowledge base for developers Managed Ops
Talk to Us

Lightsail Daily Operations Guide: Backups, Scaling, and Incident Response — Which Tasks You Should Handle Yourself

Export PDF Print-friendly layout with a copyright watermark; choose Save as PDF in the print dialog
Lightsail Daily Operations Guide: Backups, Scaling, and Incident Response — Which Tasks You Should Handle Yourself

The Bottom Line

  • Lightsail daily operations come down to three things: don't let snapshots lapse, don't overlook disks, and don't exceed the traffic quota. A 20-minute weekly inspection covers it; the hard part is not the technology, but doing it consistently for a year without missing anything.
  • It uses fixed plans, so you can only scale up or down by whole tiers; you cannot add a single block of memory. If the 5-minute average CPU exceeds 60% for two consecutive weeks, or monthly traffic exceeds 70% of the plan quota for two consecutive months, plan to migrate to EC2; don't keep upgrading to a higher plan.
  • What is worth outsourcing is "7×24 alert response + quarterly recovery drills + periodic tasks such as certificates, renewals, and disks"; business releases, SQL tuning, and architecture decisions stay in your own hands.

Monitoring and Alerting: The Dashboard Only Gives CPU and Traffic; Fill In the Rest Yourself

The Metrics page in the Lightsail console shows four categories: CPU utilization, network in/out, burst capacity headroom, and status checks. Alarms also support only these items, with email and SMS as notification channels—if you want to push to WeCom or DingTalk, you have to write your own polling relay, or use a script inside the instance to call a webhook directly.

Use the CLI to fetch historical metrics. Note that Lightsail commands must explicitly include --region; omitting it causes an error. This was a small pitfall I hit at the very beginning:

aws lightsail get-instance-metric 
  --instance-name wp-prod-01 
  --metric-name CPUUtilization 
  --period 300 
  --start-time 2026-03-01T00:00:00Z 
  --end-time 2026-03-01T06:00:00Z 
  --statistics Maximum 
  --unit Percent 
  --region ap-northeast-1


The response snippet looks like this; the jump to 99.3 is the starting point of the incident I'll discuss later:

```json
{
    "metricName": "CPUUtilization",

Sources

Leave a Reply

Your email address will not be published. Required fields are marked *