Why Chaos engineering basics in Microservices? - Purpose & Use Cases

The Big Idea

What if breaking your system on purpose could actually make it unbreakable?

The Scenario

Imagine running a busy online store with many small services talking to each other. When one service breaks, you only find out when customers complain or the whole site crashes.

The Problem

Checking each service manually for problems is slow and misses hidden issues. You can't predict how failures spread or how your system reacts under stress. This leads to surprise outages and unhappy users.

The Solution

Chaos engineering lets you safely create small failures on purpose to see how your system behaves. This helps find weak spots before real problems happen, making your system stronger and more reliable.

Before vs After

✗ Before

Wait for errors to happen, then fix them one by one.

✓ After

Inject failures automatically and watch system responses to improve resilience.

What It Enables

It enables building systems that stay strong and keep working even when parts fail unexpectedly.

Real Life Example

Netflix uses chaos engineering to randomly shut down servers and services to ensure their streaming never stops, even if something breaks.

Key Takeaways

Manual checks miss hidden failure points.

Chaos engineering tests failures proactively.

It builds confidence in system reliability.

Practice

1. What is the main goal of chaos engineering in microservices?

easy

A. To reduce the number of developers needed

B. To increase the number of microservices in a system

C. To find and fix weaknesses before real failures occur

D. To speed up the deployment process

Solution

1. Step 1: Understand chaos engineering purpose

   Chaos engineering is about testing systems by intentionally causing failures to find weaknesses.

2. Step 2: Identify the main goal

   The goal is to find and fix weaknesses before they cause real problems in production.

3. Final Answer:

   To find and fix weaknesses before real failures occur -> Option C

4. Quick Check:

   Chaos engineering goal = Find and fix weaknesses [OK]

Hint: Chaos engineering tests failures to improve system stability [OK]

Common Mistakes:

• Thinking chaos engineering increases microservices count
• Confusing chaos engineering with deployment speedup
• Assuming chaos engineering reduces developer count

2. Which of the following is a correct way to start chaos engineering experiments?

easy

A. Start with complex multi-service failures immediately

B. Begin with simple, controlled failure tests

C. Run chaos tests only after a system crash

D. Avoid monitoring during chaos experiments

Solution

1. Step 1: Review best practice for chaos experiments

   Best practice is to start small with simple, controlled failures to understand system behavior.

2. Step 2: Identify the correct starting approach

   Starting with simple tests helps safely learn and improve system resilience gradually.

3. Final Answer:

   Begin with simple, controlled failure tests -> Option B

4. Quick Check:

   Start chaos with simple tests = Begin with simple, controlled failure tests [OK]

Hint: Start chaos tests simple and controlled, not complex [OK]

Common Mistakes:

• Starting with complex failures too soon
• Running chaos only after failures happen
• Ignoring monitoring during tests

3. Consider a microservice system where a chaos experiment randomly kills one instance every 5 minutes. What is the expected immediate effect on system availability?

medium

A. System availability remains stable if redundancy exists

B. System availability drops to zero immediately

C. System crashes permanently after first kill

D. System automatically scales down instances

Solution

1. Step 1: Analyze the chaos experiment impact

   Killing one instance every 5 minutes tests resilience but does not remove all instances.

2. Step 2: Consider system redundancy

   If the system has redundant instances, killing one does not reduce availability immediately.

3. Final Answer:

   System availability remains stable if redundancy exists -> Option A

4. Quick Check:

   Redundancy keeps availability stable during chaos [OK]

Hint: Redundancy keeps system available despite instance failures [OK]

Common Mistakes:

• Assuming system crashes immediately after one instance killed
• Thinking availability drops to zero instantly
• Believing system scales down automatically

4. A chaos experiment script intended to shut down a microservice instance sometimes fails silently without stopping the instance. What is the most likely cause?

medium

A. The network is too fast for the script

B. The microservice is designed to never stop

C. The chaos experiment is running on a different system

D. The script lacks proper error handling and logging

Solution

1. Step 1: Identify why script fails silently

   Silent failures usually happen when errors are not caught or logged properly.

2. Step 2: Evaluate other options

   Microservices can be stopped; network speed does not cause silent failure; running on different system would cause errors, not silent failure.

3. Final Answer:

   The script lacks proper error handling and logging -> Option D

4. Quick Check:

   Silent failure = Missing error handling [OK]

Hint: Check error handling if chaos script fails silently [OK]

Common Mistakes:

• Assuming microservice cannot be stopped
• Blaming network speed for silent failure
• Ignoring script environment mismatch

5. You want to design a chaos engineering experiment to test how your microservices handle database latency spikes. Which approach best fits this goal?

hard

A. Inject artificial latency into database calls during tests

B. Disable monitoring tools to avoid false alerts

C. Increase the number of database replicas without testing

D. Randomly kill microservice instances during peak hours

Solution

1. Step 1: Understand the goal of testing database latency spikes

   The goal is to see how microservices behave when database responses are slow.

2. Step 2: Choose the best chaos experiment approach

   Injecting artificial latency simulates slow database calls directly, matching the goal.

3. Step 3: Evaluate other options

   Killing instances tests availability, not latency; increasing replicas without testing doesn't simulate latency; disabling monitoring hides important data.

4. Final Answer:

   Inject artificial latency into database calls during tests -> Option A

5. Quick Check:

   Test latency by injecting delays = Inject artificial latency into database calls during tests [OK]

Hint: Inject delays to test latency, not kill instances [OK]

Common Mistakes:

• Confusing instance failure with latency testing
• Adding replicas without testing effects
• Turning off monitoring during chaos