Observability
Knowing what happens in a system in production: logs, metrics, traces, the key signals to watch, percentiles rather than averages, and alerts that wake people for good reasons. Every answer explained.
- 32 cards
- 3 topics
- Free
The outline, topic by topic
Each topic is weighted the way the exam weighs it.
- Logs, metrics, traces
- Signals and percentiles
- Alerts
A few cards from the deck
What is “log”?
Show the answer
A timestamped record of one specific event, with its context
Ideal to understand one specific case, but bulky: you do not count millions of requests by reading their logs one by one.
A request takes 3 seconds crossing six services. Which tool shows most directly where the time is lost?
- A distributed trace
- The overall error rate
- One server's disk usage
- The number of logged-in users
Show the answer
A distributed trace
The trace splits the request step by step with each step's duration: the longest span points to the culprit.
A timestamped record of one specific event, with its context: what is this called?
Show the answer
log
Ideal to understand one specific case, but bulky: you do not count millions of requests by reading their logs one by one.
You want to know whether the error rate rose since the last deployment. What should you look at?
- An error rate metric over time
- A single trace
- One request's log
- The source code
Show the answer
An error rate metric over time
An aggregated metric shows the trend before and after the deployment; one isolated case says nothing about the whole.
What is “metric”?
Show the answer
A numeric measure aggregated over time, like requests per second or the error rate
Lightweight and easy to chart, it feeds dashboards and alerts; it says there is a problem, rarely why.