Monitoring and alerting in production in LangChain - Mini Project: Build & Apply

Monitoring and alerting in production

📖 Scenario: You are managing a simple web service that processes user requests. To keep the service reliable, you want to monitor the number of errors occurring and alert the team if errors exceed a certain limit.

🎯 Goal: Build a basic monitoring script that tracks error counts, sets an alert threshold, checks if the threshold is exceeded, and prints an alert message.

📋 What You'll Learn

Create a dictionary to store error counts for different services

Add a threshold variable to define the alert limit

Write logic to check if any service's error count exceeds the threshold

Print an alert message if the threshold is exceeded

💡 Why This Matters

🌍 Real World

Monitoring error counts helps keep production services reliable by alerting teams to problems early.

💼 Career

DevOps engineers and site reliability engineers use monitoring and alerting to maintain system health and uptime.

1

Create error counts dictionary

Create a dictionary called error_counts with these exact entries: 'auth_service': 3, 'payment_service': 7, 'user_service': 2

LangChain

# Create the error_counts dictionary with exact entries
# Your code here

Hint

Use curly braces to create a dictionary with keys and values separated by colons.

2

Set alert threshold

Create a variable called alert_threshold and set it to 5 to define the error count limit for alerts.

LangChain

error_counts = {'auth_service': 3, 'payment_service': 7, 'user_service': 2}
# Set the alert_threshold variable to 5
# Your code here

Hint

Just assign the number 5 to the variable alert_threshold.

3

Check for alerts

Write a for loop using variables service and count to iterate over error_counts.items(). Inside the loop, write an if statement to check if count is greater than alert_threshold.

LangChain

error_counts = {'auth_service': 3, 'payment_service': 7, 'user_service': 2}
alert_threshold = 5
# Loop over error_counts and check if count exceeds alert_threshold
# Your code here

Hint

Use a for loop with service, count and an if condition comparing count to alert_threshold.

4

Print alert message

Inside the if block, write a print statement to display the alert_message variable.

LangChain

error_counts = {'auth_service': 3, 'payment_service': 7, 'user_service': 2}
alert_threshold = 5
for service, count in error_counts.items():
    if count > alert_threshold:
        alert_message = f"Alert: {service} has {count} errors!"
        # Print the alert_message
        # Your code here

Hint

Use print(alert_message) to show the alert.

Practice

1. What is the main purpose of monitoring in a production environment?

easy

A. To send immediate messages when problems happen

B. To backup data regularly

C. To deploy new features automatically

D. To watch the app's health and performance continuously

Solution

1. Step 1: Understand monitoring role

   Monitoring means watching the app's health and performance over time.

2. Step 2: Differentiate from alerting

   Alerting is about sending messages when issues occur, not continuous watching.

3. Final Answer:

   To watch the app's health and performance continuously -> Option D

4. Quick Check:

   Monitoring = watch app health [OK]

Hint: Monitoring means watching, alerting means notifying [OK]

Common Mistakes:

• Confusing monitoring with alerting
• Thinking monitoring deploys features
• Mixing monitoring with backups

2. Which of the following is the correct way to define an alert condition in a monitoring tool?

easy

A. alert every 10 minutes regardless of CPU usage

B. alert when CPU usage > 80% for 5 minutes

C. alert when CPU usage equals 50%

D. alert if CPU usage less than 80%

Solution

1. Step 1: Identify proper alert condition

   An alert should trigger when a metric exceeds a threshold for a time period, e.g., CPU usage > 80% for 5 minutes.

2. Step 2: Eliminate incorrect options

   Alerts on less than threshold or exact equals are less useful; alerting regardless of usage is noisy.

3. Final Answer:

   alert when CPU usage > 80% for 5 minutes -> Option B

4. Quick Check:

   Alert condition = threshold + duration [OK]

Hint: Alert triggers on threshold breach over time [OK]

Common Mistakes:

• Setting alerts on exact equals
• Alerting on low usage instead of high
• Alerting without condition or duration

3. Given this alert rule snippet:

if error_rate > 5% for 10 minutes then send alert

What happens if error_rate spikes to 6% for 8 minutes and then drops to 4%?

medium

A. No alert is sent because the condition duration is not met

B. An alert is sent immediately when error_rate hits 6%

C. An alert is sent after 8 minutes

D. An alert is sent after error_rate drops below 5%

Solution

1. Step 1: Understand alert duration condition

   The alert triggers only if error_rate > 5% continuously for 10 minutes.

2. Step 2: Analyze given scenario

   Error rate was above 5% for 8 minutes, which is less than 10 minutes, so alert does not trigger.

3. Final Answer:

   No alert is sent because the condition duration is not met -> Option A

4. Quick Check:

   Duration condition unmet = no alert [OK]

Hint: Alert needs full duration breach, not just spike [OK]

Common Mistakes:

• Assuming alert triggers immediately on threshold breach
• Ignoring duration requirement
• Thinking alert triggers after drop below threshold

4. You set an alert to notify your team when memory usage exceeds 90%, but no alerts are received even though memory usage is high. What is the most likely cause?

medium

A. Notification channel is not configured correctly

B. Memory usage metric is not collected

C. Alert condition threshold is set too low

D. Alert duration is set to zero

Solution

1. Step 1: Check alert condition and metric

   Memory usage is high, so condition threshold is likely correct and metric is collected.

2. Step 2: Verify notification setup

   If no alerts are received, the notification channel (email, Slack, etc.) may be misconfigured or missing.

3. Final Answer:

   Notification channel is not configured correctly -> Option A

4. Quick Check:

   No alerts + correct condition = notification issue [OK]

Hint: Check notification setup if alerts not received [OK]

Common Mistakes:

• Assuming threshold is always wrong
• Ignoring notification channel setup
• Thinking metric collection is always faulty

5. You want to monitor a LangChain app's response time and alert the team if the average response time exceeds 2 seconds over 15 minutes. Which approach best achieves this?

hard

A. Monitor only error rates and ignore response time

B. Send an alert every time a single response takes longer than 2 seconds

C. Set up a monitoring metric for response time and alert if average > 2s for 15 minutes

D. Alert if any response time is exactly 2 seconds

Solution

1. Step 1: Define monitoring metric and alert condition

   Track average response time metric over 15 minutes to smooth out spikes.

2. Step 2: Set alert on average exceeding threshold

   Alert triggers only if average response time is above 2 seconds for the full 15 minutes.

3. Final Answer:

   Set up a monitoring metric for response time and alert if average > 2s for 15 minutes -> Option C

4. Quick Check:

   Average metric + duration alert = best practice [OK]

Hint: Alert on average over time, not single spikes [OK]

Common Mistakes:

• Alerting on single slow response
• Ignoring response time monitoring
• Alerting on exact value matches