Backend interview prep

Start

Cassandra Consistency Levels Interview Questions

Master Apache Cassandra Consistency Levels with interview questions covering Eventual Consistency, Strong Consistency, ONE, QUORUM, ALL, LOCAL_QUORUM, EACH_QUORUM, ANY, SERIAL, LOCAL_SERIAL, Read/Write Consistency, and production best practices.

Introduction

One of Cassandra's most powerful features is Tunable Consistency.

Unlike traditional databases that always provide strong consistency, Cassandra allows developers to choose the consistency level for each read and write operation.

This flexibility enables applications to balance:

  • Availability
  • Performance
  • Latency
  • Data Consistency

Understanding consistency levels is one of the most frequently asked Cassandra interview topics for Backend Engineers, Solution Architects, and Database Engineers.

Cassandra Replication Architecture

flowchart LR

Client --> Coordinator

Coordinator --> Replica1

Coordinator --> Replica2

Coordinator --> Replica3

Replication Factor = 3

1. What is Consistency in Cassandra?

Answer

Consistency determines how many replica nodes must acknowledge a read or write operation before Cassandra considers it successful.

Unlike relational databases, Cassandra allows the application to choose the required consistency level.

2. What is Tunable Consistency?

Tunable Consistency allows developers to configure consistency per query.

Example

  • Strong Consistency
  • Eventual Consistency
  • Low Latency Reads
  • Highly Durable Writes

3. What is Eventual Consistency?

Eventual Consistency means all replicas will eventually contain the same data.

Immediately after a write

Replica 1

Updated

Replica 2

Pending

Replica 3

Pending

After synchronization

Replica1

Replica2

Replica3

All Same

4. What is Strong Consistency?

Strong Consistency guarantees the latest write is returned.

Formula

Read Consistency

+

Write Consistency

>

Replication Factor

Strong Consistency

flowchart LR

Write --> Replica1

Write --> Replica2

Write --> Replica3

Read --> LatestValue

5. What is Replication Factor (RF)?

Replication Factor defines how many copies of data Cassandra stores.

Example

RF = 3

Replica1

Replica2

Replica3

6. What are Cassandra Consistency Levels?

Write Consistency Levels

  • ANY
  • ONE
  • TWO
  • THREE
  • QUORUM
  • ALL
  • LOCAL_QUORUM
  • EACH_QUORUM
  • LOCAL_ONE

Read Consistency Levels

  • ONE
  • TWO
  • THREE
  • QUORUM
  • ALL
  • LOCAL_QUORUM
  • LOCAL_ONE
  • SERIAL
  • LOCAL_SERIAL

7. Explain Consistency Level ONE.

Write succeeds after

1 Replica

acknowledges.

Advantages

  • Fast
  • Low Latency

Disadvantages

  • Eventual Consistency

ONE Example

flowchart LR

Coordinator --> Replica1

Coordinator

-.->Replica2

Coordinator

-.->Replica3

8. Explain TWO.

Two replicas must acknowledge.

Higher consistency than ONE.

9. Explain THREE.

Three replicas acknowledge.

Usually used when RF ≥ 3.

10. Explain QUORUM.

Formula

(RF / 2)

+

1

Example

RF = 3

2 Replicas

must respond.

QUORUM Example

flowchart LR

Coordinator --> Replica1

Coordinator --> Replica2

Coordinator

-.->Replica3

11. Explain LOCAL_QUORUM.

Only replicas within the local data center participate.

Ideal for

  • Multi-DC Applications
  • Low Latency

12. Explain EACH_QUORUM.

Requires quorum from every data center.

Example

US-East

↓

2 Nodes

Europe

↓

2 Nodes

Used rarely due to higher latency.

13. Explain ALL.

All replicas must acknowledge.

Advantages

  • Highest Consistency

Disadvantages

  • Slow
  • Less Available

14. Explain ANY.

ANY requires only one acknowledgment.

The acknowledgment may come from:

  • Replica
  • Hint Handoff

Fastest write consistency.

15. Explain LOCAL_ONE.

Only one replica from the local data center responds.

Useful for

  • Multi-Region Applications
  • Low Latency

16. What is SERIAL?

Used for

Lightweight Transactions (LWT)

Implements Paxos protocol.

Example

INSERT ...

IF NOT EXISTS

17. What is LOCAL_SERIAL?

Same as SERIAL

Only within the local data center.

18. Difference between QUORUM and LOCAL_QUORUM?

QUORUM LOCAL_QUORUM
All Data Centers Local DC Only
Higher Latency Lower Latency
Cross-DC Local Reads/Writes

19. Difference between ONE and QUORUM?

ONE QUORUM
Faster More Consistent
Lower Latency Slightly Higher Latency
Eventual Consistency Stronger Consistency

20. Difference between QUORUM and ALL?

QUORUM ALL
Majority Every Replica
Better Availability Lower Availability
Faster Slowest

21. Explain Read Consistency.

Read Consistency determines

How many replicas must respond before returning data.

22. Explain Write Consistency.

Write Consistency determines

How many replicas acknowledge before write success.

23. Explain the Strong Consistency Formula.

Formula

Read CL

+

Write CL

>

Replication Factor

Example

RF = 3

Write QUORUM = 2

Read QUORUM = 2

2 + 2 > 3

Strong Consistency

24. Why does QUORUM provide strong consistency?

Because majority replicas participate.

Latest write is guaranteed.

25. What happens if one replica is unavailable?

Depends on consistency level.

Example

RF = 3

ONE

↓

Still succeeds

ALL

↓

Fails

26. Which consistency level is fastest?

ANY

↓

ONE

↓

LOCAL_ONE

27. Which consistency level is safest?

ALL

Highest consistency.

28. Which consistency level is most commonly used?

Most enterprise applications use

LOCAL_QUORUM

because it balances

  • Performance
  • Availability
  • Consistency

29. What consistency level does Netflix commonly prefer?

Generally

LOCAL_QUORUM

Reason

  • Multi-Region
  • High Availability
  • Strong Local Consistency

30. Real Production Example

Banking Application

Replication Factor

3

Transaction Write

LOCAL_QUORUM

Transaction Read

LOCAL_QUORUM

Benefits

  • Strong Consistency
  • Low Latency
  • Multi-DC Support

Consistency Flow

flowchart LR

Client --> Coordinator --> Replicas --> ConsistencyCheck --> Response

Enterprise Best Practices

  • Use RF = 3 for production.
  • Prefer LOCAL_QUORUM in multi-data-center environments.
  • Avoid ALL unless absolutely necessary.
  • Use ONE for analytics or less critical data.
  • Use SERIAL only for lightweight transactions.
  • Monitor replica latency.
  • Keep clocks synchronized across nodes.
  • Schedule regular repairs to reduce inconsistencies.
  • Test consistency settings under failure scenarios.

Quick Revision

Consistency Level Description
ANY Fastest Write
ONE One Replica
TWO Two Replicas
THREE Three Replicas
QUORUM Majority
LOCAL_QUORUM Majority in Local DC
EACH_QUORUM Majority in Every DC
ALL Every Replica
LOCAL_ONE One Replica in Local DC
SERIAL LWT Consistency
LOCAL_SERIAL Local LWT

Interview Tips

Interviewers commonly ask

  • What is Tunable Consistency?
  • Explain QUORUM.
  • Explain LOCAL_QUORUM.
  • Difference between ONE and QUORUM.
  • Explain Strong Consistency Formula.
  • Which consistency level would you choose for banking?
  • Which consistency level provides the fastest writes?
  • What happens when a replica is down?
  • Explain SERIAL consistency.
  • Why doesn't Cassandra always use strong consistency?

Always explain the trade-off between Consistency, Availability, and Latency, and justify your consistency level based on the application's business requirements.

Summary

Cassandra's tunable consistency model gives applications the flexibility to choose the appropriate balance between consistency, availability, and performance. By selecting consistency levels such as ONE, QUORUM, LOCAL_QUORUM, or ALL, developers can optimize their systems for different workloads and failure scenarios.

Understanding replication factor, read and write consistency, the strong consistency formula, and the behavior of each consistency level is essential for designing reliable, scalable, and highly available Cassandra deployments in enterprise environments.