Cassandra Interview Questions & Answers (2025) | JaganInfo

Apache Cassandra is a distributed NoSQL database designed to handle large volumes of data with high availability and no single point of failure.

Cassandra is a wide-column store NoSQL database.

It was originally developed by Facebook to power its Inbox search feature and later open-sourced.

Keyspace is the top-level data container in Cassandra, similar to a database in RDBMS, defining replication settings.

A primary key uniquely identifies rows in a table, consisting of a partition key and optional clustering columns.

A column family is a structure to store data in columns and rows, similar to a table but more flexible in schema.

Cassandra provides tunable consistency and lightweight transactions (partial ACID), but not full ACID for multi-row operations.

A node is an individual instance of Cassandra running in a cluster, storing part of the data.

Replication is storing copies of data on multiple nodes to ensure reliability and high availability.

Cassandra uses CQL (Cassandra Query Language), similar in syntax to SQL but without joins.

A cluster is a collection of nodes arranged to distribute data evenly and provide redundancy.

It defines how many replica nodes must acknowledge a read/write operation before it’s considered successful.

Partition key determines data distribution; clustering key orders rows within a partition.

Sorted String Table is an immutable data file where Cassandra stores data after memtable flush.

An in-memory data structure where writes are first stored before flushing to disk as SSTables.

Compaction merges SSTables, discards old data, and enhances read efficiency.

A mechanism to temporarily store hints for failed nodes and deliver missed writes when they recover.

A background process that fixes inconsistencies between replicas during a read request.

The node that receives a client request and forwards it to appropriate replica nodes for action.

A peer-to-peer communication method used by Cassandra nodes to share information about cluster state.

Yes, it supports multi–data center replication with configurable strategies.

Snitches determine network topology to route requests and place replicas efficiently.

It allows clients to choose consistency level per query balancing speed and data accuracy.

No joins, no subqueries, requires using partition key for most queries.

Guarantee that, in the absence of further writes, all replicas will eventually converge to the same value.

Modeling is query-driven, denormalized, storing data to suit access patterns.

Automatically updated view of data based on an existing table with different primary key.

Using JMX metrics, OpsCenter, Prometheus, Grafana.

Yes, it supports client-to-node, node-to-node, and at-rest encryption.

Examples: ONE, QUORUM, ALL — number of replicas that must respond for read to succeed.

Through replication, hinted handoff, read repair, and anti-entropy repair processes to ensure consistency.

Write goes to commit log, then memtable; later, memtable flushes to SSTable on disk.

Coordinator queries replicas, merges results from memtable, row cache, and SSTables, returns latest data based on timestamp.

Vnodes allow each node to own multiple token ranges, improving load distribution and rebalance speed.

Adding nodes automatically redistributes data without downtime.

Process to synchronize data between replicas to fix inconsistencies.

They use Paxos consensus to provide compare-and-set operations for conditional updates.

The time for which hints are stored for a down node before they are discarded.

Choose keys that evenly distribute data to avoid hotspots, considering query patterns.

A performance feature that retries reads on another replica if the original is slow to respond.

Stores entire rows in memory to speed up reads, but increases memory usage.

Replication strategy is configured to replicate specified number of copies to each data center.

It’s used to replay unflushed writes during restart after a crash to prevent data loss.

Markers for deleted data that remain until compaction permanently removes them.

Enable authentication, encryption, set proper file permissions, and limit network access.

Seed nodes are contact points for new or restarted nodes to join the cluster’s gossip ring.

Use snapshots and incremental backups; restore by loading SSTables from backups.