Latency monitoring per step in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

💻 Command Output

intermediate

Interpreting latency output from a step monitoring tool

You run a latency monitoring command on a pipeline step and get this output:

Step: DataFetch
Latency: 120ms
Status: Success

What does this output tell you?

A The DataFetch step has a latency of 120 microseconds and failed.

B The DataFetch step failed after 120 seconds.

C The entire pipeline took 120 milliseconds to run.

D The DataFetch step took 120 milliseconds to complete successfully.

💡 Hint

Latency is usually measured in milliseconds and indicates duration.

Latency output shows how long a step took. Here, 120ms means 120 milliseconds, and status Success means it completed without error.

❓ Configuration

intermediate

Configuring latency monitoring for a pipeline step

You want to configure latency monitoring for a step named 'ProcessData' using a YAML config. Which snippet correctly sets a threshold of 200ms to alert if exceeded?

A

steps:
  - name: ProcessData
    latency:
      max_time: 0.2
      alert_on_exceed: true

B

steps:
  - name: ProcessData
    latency_monitor:
      alert_threshold: 200ms
      enabled: yes

C

steps:
  - name: ProcessData
    latency_monitor:
      threshold_ms: 200
      alert: true

D

steps:
  - name: ProcessData
    monitor_latency:
      threshold: 200
      alert_enabled: true

💡 Hint

Look for correct key names and units in milliseconds.

Option C uses correct keys 'latency_monitor', 'threshold_ms', and 'alert' with boolean true, matching common config patterns.

❓ Troubleshoot

advanced

Diagnosing missing latency data in monitoring logs

You notice that latency data for step 'Analyze' is missing in your monitoring logs. Which is the most likely cause?

A The 'Analyze' step completed too quickly to record latency.

B The latency monitoring agent was not enabled for the 'Analyze' step.

C The monitoring logs are corrupted and missing all data.

D The 'Analyze' step is running on a different server without network access.

💡 Hint

Check if monitoring is enabled per step.

Latency data is only collected if the monitoring agent is enabled for that step. If missing, likely it was not enabled.

🔀 Workflow

advanced

Optimizing pipeline based on latency monitoring data

Your latency monitoring shows step 'Transform' consistently takes 500ms, which is double the threshold. What is the best next step?

A Investigate the 'Transform' step code and resources to identify bottlenecks.

B Ignore the latency since the step still completes successfully.

C Increase the latency threshold to 600ms to avoid alerts.

D Disable latency monitoring for the 'Transform' step to reduce overhead.

💡 Hint

High latency should trigger investigation to improve performance.

Consistent high latency means the step is slow. Investigating code and resources helps find and fix bottlenecks.

✅ Best Practice

expert

Implementing latency monitoring with minimal performance impact

Which approach best balances accurate latency monitoring per step with minimal impact on pipeline performance?

A Use asynchronous logging of latency data after step completion to avoid blocking the step.

B Add synchronous latency measurement code inside each step, logging before and after execution.

C Collect latency data only once per pipeline run to reduce overhead.

D Disable latency monitoring during peak hours to improve performance.

💡 Hint

Consider how monitoring affects step execution time.

Asynchronous logging records latency without delaying the step, minimizing performance impact while keeping data accurate.