पाठ 33 / 35

Common Interview Questions

यह पाठ अभी हिंदी में अनुवादित नहीं है, इसलिए अंग्रेज़ी में दिखाया जा रहा है।

Answer frequently asked observability interview questions with clear, structured explanations and examples.

लेखक: · 15 साल से वेब डेवलपर

Questions you are likely to hear (part 1)

"What is the difference between monitoring and observability?" Monitoring checks predefined conditions; observability is the ability to ask new questions of telemetry. Give an example: a dashboard shows latency is up; observability lets you slice by region and customer tier to find that only one replica is slow.

"Counter vs gauge?" A counter only increases and you graph its rate; a gauge goes up and down and you graph its value. Mention counter resets and that rate() handles them.

"Why not use average latency?" Latency is skewed; averages hide the tail. Use percentiles from histograms, and aggregate buckets before computing quantiles.

"What is high cardinality and why is it a problem?" Many unique label combinations create many time series, consuming memory and slowing queries. Example: a user_id label. Fix by moving IDs to logs or traces.

Structure every answer

Interviewers reward structure: define the term in one sentence, explain why it matters, give a concrete example from a system you know, and finish with a trade-off or pitfall. Ninety seconds is a good target for each answer.

Questions you are likely to hear (part 2)

"How does distributed tracing work?" Each request gets a trace ID; every operation records a span with its parent ID; context crosses services in the W3C traceparent header; a backend rebuilds the tree. Mention sampling and why it is needed.

"Head vs tail sampling?" Head decides at the start and is cheap but blind to outcomes; tail decides after the trace completes and can keep errors and slow traces but needs buffering in the collector.

"How would you design alerting for a new service?" Define SLIs (success ratio, latency), set an SLO, alert on multi-window burn rate with runbooks, route pages and tickets separately, and review alert quality regularly.

"An endpoint is slow, how do you investigate?" Confirm with RED metrics and percentiles, narrow by route and instance, open exemplar or sample traces to find the slow span, read logs for that span, then check saturation of the dependency involved.

त्वरित जाँच: An interviewer asks how you would avoid paging for a 30-second error spike while still catching a real outage quickly. What is the strongest answer?

  • Remove error alerts and rely on user complaints
  • Alert when any single request fails
  • Use a multi-window burn-rate alert: a long window for significance plus a short window to confirm it is ongoing
  • Average the error rate over 30 days and alert monthly
Answer

Use a multi-window burn-rate alert: a long window for significance plus a short window to confirm it is ongoing — Multi-window burn-rate alerts ignore brief spikes yet page quickly for serious, ongoing budget consumption.