QA finding
A monitored address was down, slow or answered wrongly
QA Error Severity: High
Part of the audit Uptime and response time. Finding id: qa.uptimeIncidents.
What it costs you
While the address fails, visitors cannot use the site and orders or enquiries are lost.
What to do
Compare the times with your server logs and your hosting provider's status page to find the cause.
What it checks
- Site not reachable
- Site responds too slowly
- Unexpected response code
- Security certificate not valid
- Expected content not found
How to fix it
Renew or repair the certificate
- Open the address in a browser and read the certificate warning: expired, wrong name or incomplete chain.
- Renew the certificate, or install one that covers this host name together with its intermediate certificates.
- If renewal is automatic, find why it did not run.
How to check: The address opens in a browser without a warning, and the next checks from every location pass.
Find why the address answers with an error status
- Request the address yourself and read the status.
- 401 or 403: look for an access rule or a firewall that blocks the monitor or visitors. 404 or 410: the page was removed or moved; restore it or monitor its new address.
- Compare the start time with deployments and configuration changes.
How to check: The next checks from every location answer 200.
Find what the page showed
- Open the page. An error page or an empty template can answer 200 too.
- If the page is fine and its text was changed on purpose, update the expected text of the monitor.
- If not, look in the application log at the start time for the failure behind it.
How to check: The next checks find the expected text.
Find what slows the answers
- Plot the response times of the checks over the day.
- Compare slow hours with traffic peaks, backups and other scheduled jobs.
- Check whether pages are served from a cache, and whether one location is much slower than the others.
How to check: The average response time of the next reporting period is under the limit.
Find the cause of the outage
- Give the start and end time of the incident to whoever runs the server.
- Look in the server and application logs at those times for restarts, full disks, exhausted connections or deployments.
- Check the status page of the hosting provider for the same period.
How to check: The next checks from every location answer 200, with the expected text, within the limit.
A real case
Researchers at Uptime Institute found that over 60% of the outages in their 2022 survey caused at least $100,000 in total losses, up from 39% in 2019, and 15% cost more than $1 million.
Thresholds
| Name | Default | Meaning |
|---|---|---|
min_locations | 2 | An outage seen from fewer probe locations than this is treated as a network problem on our side and goes to review. |
max_response_ms | 2000 | Average response time of successful checks over the reporting period, in milliseconds, above which the address is returned as slow. |