APPLICATION PERFORMANCE MONITORING

APM: Magic Injections and Massive Bills.

Visualization: 1. Code-Level Visibility (Why APM?)

Metrics

'The DB is slow'

Logs

'Timeout on query'

APM

Deep Introspection

Flame Graph

Line 42 takes 3s

DOWNLOADING CONCEPT: 1. CODE-LEVEL VISIBILITY (WHY APM?)...

↳Metrics tell you *that* a system is failing.
↳Logs tell you *what* error was thrown.
↳But when a specific API endpoint is inexplicably taking 4 seconds to respond, and the database looks perfectly healthy, neither metrics nor logs can easily tell you *why*.

Application Performance Monitoring (APM)

↳bridges this gap.
↳APM tools give you code-level visibility.
↳Instead of guessing, you open the APM dashboard and look at a Flame Graph that visually proves: 'The request took 4 seconds because the JSON serialization library on line 42 burned 3.8 seconds of CPU time'.
↳It points you directly to the offending line of code.

▲ LIVE INCIDENTThe APM dashboard shows a method processData() taking 4 seconds, but there are no database calls or HTTP requests inside it.

sre@arena:~/apm

sre@arena:~$ jstack <PID> > thread_dump.txt && grep -A 10 "processData" thread_dump.txt