My site is 'up' but painfully slow — should monitoring catch that?
Q: Uptime says green but pages take 8 seconds. Users are leaving. Shouldn't my monitor flag slow as a problem?
A: It should, and a reader's data proves the cost of ignoring it. They alerted only on hard failures. Their API was technically 'up' but p95 response time crept from 400ms to 4.2 seconds over three weeks as a query degraded. No page ever fired. Conversion quietly dropped about 19% before anyone connected the dots.
They added a latency threshold: alert if response time exceeds 1.5 seconds for 5 consecutive checks. The next slow-degradation was caught while p95 was still 1.6s — long before users abandoned in droves. Root cause was a missing database index, fixed in an afternoon.
The follow-up: alert on a sustained breach, not a single slow check. One slow probe is noise; five in a row is a trend. And measure p95 or p99, not the average — averages hide the users who are actually suffering.
Got a question? Drop it in the comments.
Pingback Clinic
@PingbackClinic
My site is 'up' but painfully slow — should monitoring catch that?
Этот пост опубликован в Telegram-канале Pingback Clinic. Подписаться можно по ссылке: @PingbackClinic.