Apache Cassandra® versions: What’s new and how to make the move

What are the latest Apache Cassandra versions?

The following table summarizes the latest versions of Apache Cassandra. The project aims to deliver one major release per year, with patch releases issued based on the urgency or volume of bug fixes.

New features in recent Cassandra versions

Cassandra 5.0

Cassandra 5.0 expands functionality for performance, analytics, security, and AI/ML workloads.

• Storage-attached indexes (SAI): Provides scalable, high-throughput indexing across data types, including vectors, with efficient index streaming.
• Vector search: Enables similarity search for AI use cases using dense indexing techniques on high-dimensional data.
• Unified compaction strategy: Combines multiple compaction approaches to reduce SSTable size and improve read and write efficiency.
• Trie-based memtables and SSTables: Uses trie data structures to improve storage efficiency and optimize read and write performance.
• New mathematical CQL functions: Adds functions such as abs, exp, log, log10, round, and collection-level aggregations like count, max, min, sum, and avg.
• Dynamic data masking: Allows runtime masking of sensitive data to enforce privacy and compliance.
• Stability and testing improvements: Expands testing coverage and improves overall system reliability.

Cassandra 4.0

Cassandra 4.0 established a new stability baseline and introduced several core improvements across runtime, logging, and data distribution.

• Java 11 support: Full support for running Cassandra on Java 11.
• Virtual tables: System and operational data can be queried through virtual tables using CQL.
• Audit logging: Tracks database activity for compliance and security auditing.
• Full query logging: Captures executed queries for troubleshooting and workload analysis.
• Improved internode messaging: Enhances reliability of communication between cluster nodes.
• Improved streaming: Increases robustness and reliability of data streaming between nodes.
• Transient replication: Allows selective replication to reduce storage overhead while maintaining availability.

Cassandra 4.1

Cassandra 4.1 focused on pluggability, operational control, performance improvements, and stronger security.

• Paxos improvements: Optimizes Lightweight Transactions (LWTs), reduces round trips, improves latency, and maintains linearizable consistency across range movements.
• Improved configuration framework: Redesigns cassandra.yaml with clearer syntax, standardized naming, and typed configuration values while preserving backward compatibility.
• New CQL features: Adds grouping by time range, new conditional clauses (CONTAINS, CONTAINS KEY), and support for IF EXISTS and IF NOT EXISTS in ALTER statements.
• Security enhancements: Adds authentication plugin support in CQLSH, PEM-based TLS key material, and a tool for generating hashed passwords for role management.
• Pluggable memtable implementation: Enables custom memtable implementations and per-table configuration.
• Pluggable SSL context creation: Allows custom handling of TLS artifacts to meet enterprise security requirements.
• Pluggable schema management: Enables schema storage outside local system tables for external integration.
• Guardrails framework: Enforces soft and hard limits to prevent misconfiguration and performance issues.
• Partition denylisting: Lets operators isolate problematic partitions that affect cluster performance.
• SSTable identifiers: Introduces globally unique identifiers to simplify backup and restore operations.
• Cluster and code simulator: Provides large-scale simulations to validate consensus behavior and correctness.

How to upgrade Cassandra from 4.x to 5.x

Cassandra uses semantic versioning in the format MAJOR.MINOR.PATCH, where each part signals the type of change. Major versions introduce breaking changes, minor versions add backward-compatible features, and patch versions contain bug fixes. Understanding this helps determine upgrade risk and planning effort.

Upgrade types and planning

Major upgrades (e.g., 4.x → 5.x) require careful validation. These changes can introduce incompatibilities, so the upgrade path should be reviewed in advance. It is strongly recommended to test the upgrade on a non-production cluster before applying it to production. This helps identify issues with drivers, schemas, or application behavior.

Minor and patch upgrades (e.g., 4.1.0 → 4.1.2) are lower risk. They typically include bug fixes and small improvements. Even though they are safer, they should still be validated in staging environments when possible.

Typical upgrade workflow

In managed environments, upgrades are usually orchestrated with minimal or no downtime by upgrading nodes in stages (e.g., upgrading the cluster rack by rack). A standard process includes:

1. Backup the cluster: A full backup is taken to ensure recovery is possible if issues occur.
2. Upgrade one rack at a time: Cassandra is stopped on all nodes in a rack, upgraded, and restarted. This limits impact and maintains cluster availability.
3. Application verification: After each rack upgrade, the application is checked to confirm it is functioning correctly before proceeding.
4. Repeat for remaining racks: The process continues rack by rack until all nodes run the new version.
5. Run SSTable upgrades: A background process upgrades on-disk data formats (upgradesstables). This step can take days depending on data size and increases CPU usage. While Cassandra can run with mixed formats, performance may degrade until this step is complete.
6. Final validation: A final check ensures the cluster and applications are stable after the upgrade.

Managed Cassandra 5.0 on Instaclustr

Cassandra 5.0 introduces several new features focused on performance, scalability, and security, many of which are supported on the Instaclustr managed platform, including:

• Storage-attached indexes (SAI): Enable highly efficient, column-level indexing across diverse data types, supporting fast, low-latency searches and advanced use cases such as vector search for AI and similarity-based querying.
• Unified compaction strategy: Combines leveled, tiered, and time-windowed compaction into a single approach, producing smaller SSTables that improve I/O performance, reduce storage overhead, and simplify tuning and operations.
• Trie-based memtables and SSTables: Improve read/write throughput and memory efficiency, along with CQL enhancements that add new scalar and aggregation functions such as abs, sum, avg, and log10 to expand analytical capabilities.
• Dynamic data masking: Strengthens security by protecting sensitive fields in real time without modifying the underlying data, while also delivering general improvements in stability, test coverage, monitoring, and virtual table functionality.

These features make Cassandra 5.0 on Instaclustr a strong choice for teams needing scalable, secure, and high-performance data infrastructure.

Getting started

If you are not an existing customer, click here to sign up for a free trial to experience seamless deployments of Apache Cassandra on the Instaclustr Managed Platform. Once you have signed up, visit our documentation site to learn how to spin a Cassandra cluster with just a few clicks.

If you have any further queries regarding this release, please contact Instaclustr Support.