What's new in MinIO AIStor RELEASE.2026-02-02?

Asked by muratkars Answered by muratkars February 1, 2026
0 views

This major release transforms MinIO AIStor into a complete data lakehouse foundation with production-ready S3 Tables/Iceberg Catalog and full Delta Sharing protocol implementation. Combined with infrastructure improvements for dynamic pool management and enhanced monitoring, this release enables enterprise analytics workloads at scale.

Answer

RELEASE.2026-02-02 is a landmark release focused on data lakehouse enablement. The S3 Tables/Iceberg Catalog reaches production maturity with Views API, hierarchical namespaces, and significant performance optimizations. The complete Delta Sharing protocol implementation enables secure cross-organization data collaboration. Infrastructure improvements include dynamic pool hot-reload, S.M.A.R.T. disk monitoring, and extended erasure coding support.

EnvironmentRecommendation
Data lakehouse/analytics deploymentsUpgrade immediately
Production deploymentsUpgrade at next maintenance window
All other environmentsPlan upgrade within 30 days

S3 Tables/Iceberg Catalog — Production Ready

Iceberg Views API

What

Full implementation of Iceberg Views API enabling logical views over table data:

  • Create, load, replace, and drop view operations
  • Schema evolution support for views
  • View metadata persistence with versioning
  • Cross-table views spanning multiple Iceberg tables

Availability

Available in this release.

Why

Views are essential for analytics workflows—they enable logical abstractions over physical tables, simplify complex queries, and provide access control boundaries. The Iceberg specification defines Views as first-class citizens alongside Tables, and production analytics platforms expect this capability.

What This Means for Customers

  • Simplified analytics — Create reusable query abstractions without data duplication
  • Access control — Grant access to views without exposing underlying tables
  • Spark/Trino compatibility — Views work seamlessly with popular query engines
  • No additional cost — Views are metadata-only, consuming minimal storage

Hierarchical Namespaces

What

Support for multi-level namespace hierarchies enabling complex organizational structures:

catalog.database.schema.table
catalog.team.project.dataset
  • Nested namespaces with arbitrary depth
  • Namespace-level permissions for fine-grained access control
  • Path-based organization matching enterprise data governance patterns

Availability

Available in this release.

Why

Enterprise data platforms organize data hierarchically—by business unit, project, environment, and dataset. Flat namespace models force awkward naming conventions and complicate access control. Hierarchical namespaces align with how organizations actually structure their data.

What This Means for Customers

  • Enterprise data organization — Model your data catalog structure naturally
  • Simplified governance — Apply policies at namespace boundaries
  • Multi-tenant support — Isolate tenants with namespace hierarchies
  • Migration path — Match existing Hive/Glue catalog structures

Performance Optimizations

What

Major performance improvements for Iceberg catalog operations:

OptimizationImprovement
Msgpack storage formatFaster serialization than JSON
Namespace sharding16 separate files eliminate write contention
Shard-based paginationReduced memory usage for large catalogs
Metadata cachingMinimized redundant lookups
Per-warehouse locksReplaced global lock, reduced contention

Availability

Available in this release. Msgpack migration is automatic and backward-compatible.

Why

Production Iceberg catalogs can contain millions of tables across thousands of namespaces. The previous single-file storage model created contention during concurrent operations. Sharding distributes load and eliminates bottlenecks.

What This Means for Customers

  • Higher concurrency — More simultaneous catalog operations
  • Lower latency — Faster table creation and metadata lookups
  • Larger scale — Support for enterprise-scale catalogs
  • No action required — Optimizations apply automatically

Transaction Framework Improvements

What

Enhanced transaction handling for catalog operations:

  • Unified recovery framework consolidates recovery code paths
  • Atomic RenameTable with persistent transaction logs
  • Automatic recovery on server restart

Availability

Available in this release.

Why

Catalog operations must be atomic—partial failures corrupt metadata state. The unified recovery framework ensures consistent state recovery regardless of failure timing.

What This Means for Customers

  • Data integrity — No partial or corrupted catalog state
  • Crash recovery — Automatic recovery after unexpected restarts
  • Operational confidence — Reliable catalog operations under load

Table Authorization System

What

New TableAuthSys provides per-table and per-view authorization:

  • Fine-grained permissions at table/view level
  • Authorization caching reduces IAM policy evaluation overhead
  • Integration with existing IAM policies and roles

Availability

Available in this release.

Why

Catalog-level authorization is insufficient for production analytics—teams need table-level access control. Repeated IAM policy evaluations created performance overhead that caching eliminates.

What This Means for Customers

  • Granular access control — Permissions at table level, not just bucket level
  • Faster authorization — Cached evaluations reduce latency
  • Compliance ready — Audit-friendly access control model

Delta Sharing Protocol — Complete Implementation

What

Full implementation of Delta Sharing v1.0 protocol for secure cross-organization data sharing:

CapabilityDescription
Complete REST APIFull specification compliance
Bearer token authStandard authentication mechanism
Delta Lake integrationNative log parsing and format support
Predicate pushdownQuery optimization for efficient data transfer
Databricks compatibilityWorks with Databricks sharing clients
OAuth 2.0 supportClient credentials flow for automation

Scale: 19,740+ lines of code across 62 files—a complete, production-ready implementation.

Availability

Available in this release.

Why

Organizations need to share data securely with partners, customers, and across business units without copying data or exposing infrastructure. Delta Sharing provides an open protocol for this use case, and MinIO’s implementation enables sharing directly from AIStor without additional infrastructure.

What This Means for Customers

  • Secure data sharing — Share data without exposing storage infrastructure
  • No data copying — Recipients query data in place
  • Cross-platform — Works with any Delta Sharing-compatible client
  • Partner collaboration — Enable secure data exchange with external organizations
  • Cost reduction — Eliminate data duplication for sharing use cases

Infrastructure Improvements

Pool Hot-Reload

What

Dynamically add or remove storage pools without server restart:

  • SIGHUP signal triggers pool configuration reload
  • Zero-downtime capacity expansion
  • Live pool removal for decommissioning

Availability

Available in this release.

Why

Production deployments shouldn’t require restarts for capacity changes. Hot-reload enables truly continuous operations during infrastructure modifications.

What This Means for Customers

  • Zero downtime — Add capacity without service interruption
  • Operational flexibility — Respond to capacity needs immediately
  • Simplified maintenance — No restart coordination required

S.M.A.R.T. Disk Monitoring

What

Pure Go implementation of disk health monitoring:

  • NVMe drives — Health status, temperature, wear leveling
  • SATA drives — S.M.A.R.T. attribute monitoring
  • Predictive alerts — Early warning of drive degradation
  • No external dependencies — Built into MinIO binary

Availability

Available in this release.

Why

Proactive disk monitoring prevents data loss by identifying failing drives before they fail completely. Native integration eliminates dependency on external monitoring tools.

What This Means for Customers

  • Proactive maintenance — Replace drives before failure
  • Reduced risk — Early warning of degradation
  • Simplified operations — No external monitoring tools required
  • Observability — Integrated with MinIO metrics

Thread Pressure Monitoring

What

Automatic backpressure when server reaches critical thread exhaustion:

  • HTTP 429 response at ≥85% thread utilization
  • Prevents cascading failures from resource exhaustion
  • Client-side retry signals enable graceful degradation

Availability

Available in this release.

Why

Under extreme load, accepting more requests than can be processed causes cascading failures. Returning 429 early allows clients to back off and retry, maintaining overall system stability.

What This Means for Customers

  • System stability — Prevents runaway resource exhaustion
  • Graceful degradation — Clients receive retry signals
  • Better SLAs — Predictable behavior under overload

Extended Erasure Coding

What

Support for up to 32 drives per erasure set (default remains 16):

  • Environment variable: MINIO_ERASURE_SET_DRIVE_COUNT (17-32)
  • Higher fault tolerance with more parity drives
  • Larger erasure sets for high-density configurations

Availability

Available in this release. Opt-in via environment variable.

Why

High-density storage servers with 36+ drives benefit from larger erasure sets. Extended support enables optimal configurations for modern hardware.

What This Means for Customers

  • Hardware flexibility — Optimize for high-density servers
  • Higher durability — More parity drives if needed
  • Cost efficiency — Better storage utilization with large drive counts

Async Identity Provider Startup

What

Server starts successfully when LDAP/OIDC providers are temporarily unavailable:

  • Deferred authentication until provider becomes available
  • Graceful degradation for identity provider outages
  • Automatic reconnection when provider recovers

Availability

Available in this release.

Why

Identity provider dependencies shouldn’t prevent MinIO startup. Transient LDAP/OIDC outages during maintenance windows shouldn’t cascade to storage unavailability.

What This Means for Customers

  • Improved availability — Start despite IdP outages
  • Operational resilience — Survive IdP maintenance windows
  • Reduced dependencies — Looser coupling to identity infrastructure

Performance Optimizations

ListObjects Improvements

What

Heap-based k-way merge algorithm for ListObjects operations:

  • Reduced memory allocation during large listings
  • Faster merge operations across erasure sets
  • Improved pagination performance

Availability

Available in this release.

Why

ListObjects with millions of keys requires efficient merge algorithms. Heap-based k-way merge is optimal for this workload.

What This Means for Customers

  • Faster listings — Reduced latency for large buckets
  • Lower memory — More efficient resource utilization

PutObject Temp Directory Sharding

What

16-way sharding of temporary directories for uploads:

  • Reduced filesystem contention during concurrent uploads
  • Better parallelism for high-throughput workloads

Availability

Available in this release.

Why

Single temp directory creates filesystem contention during high-concurrency uploads. Sharding distributes load across multiple directories.

What This Means for Customers

  • Higher throughput — More concurrent uploads without contention
  • Better performance — Especially for small object workloads

Regex Precompilation

What

Precompiled regular expressions eliminate ~1200ns/op overhead per request.

Availability

Available in this release.

Why

Request-path regex compilation adds unnecessary latency. Precompilation moves this cost to startup.

What This Means for Customers

  • Lower latency — Reduced per-request overhead
  • Consistent performance — No compilation jitter

Experimental Features

Server-Side Logging (Tech Preview)

What

Persistent API, Error, and Audit event logging with multiple targets:

  • Internal disk recording for local persistence
  • External webhook targets for log aggregation
  • Kafka integration for streaming pipelines
  • Configurable batching with persistent queue recovery

Configuration options:

  • log_api_internal, log_error_internal, log_audit_internal
  • Webhook and Kafka target variants

Availability

Tech preview. Opt-in, disabled by default.

Why

Centralized logging infrastructure isn’t always available or appropriate. Server-side logging provides persistence guarantees for audit requirements.

What This Means for Customers

  • Audit compliance — Persistent logs for regulatory requirements
  • Operational visibility — Capture events even without external logging
  • Flexibility — Multiple target options for different architectures

Security Updates

What

  • CVE fixes from Go 1.25.6 security updates
  • Environment variable redaction for sensitive data in logs
  • TLS certificate error message improvements

Availability

Available in this release.

Why

Security updates must be applied promptly. Sensitive data redaction prevents credential exposure in logs.

What This Means for Customers

  • Security compliance — Latest CVE patches
  • Credential protection — No sensitive data in logs
  • Easier debugging — Improved TLS error messages

Bug Fixes

IssueFix
Race conditions in WaitGroup, metrics, lock filesEliminated data races
Replication status correctionsAccurate replication state reporting
Disabled replication target filteringProper target state handling
Decommission metadata handlingObject readability verification
JWKS pull fallbackImproved OIDC key retrieval reliability
OpenID access key managementCorrect key lifecycle handling

Breaking Changes

ChangeImpactMigration
DROP TABLE defaults purgeRequested to trueSpark compatibility workaroundExplicit purgeRequested=false if needed
Tables metadata uses msgpack formatStorage format changeAutomatic migration, backward-compatible
v3 metrics returns up to 1000 bucketsPreviously limited to 10May affect metric consumers expecting fewer buckets

New Configuration Options

OptionDefaultDescription
MINIO_ERASURE_SET_DRIVE_COUNT16Extended range: 17-32 drives per set
MINIO_REPLICATION_EVENT_THRESHOLD15mReplication event processing threshold
PurgeOnDelete: sparkPer-bucket versioning config for Spark compatibility

Summary

RELEASE.2026-02-02 establishes MinIO AIStor as a complete data lakehouse foundation:

  • S3 Tables/Iceberg Catalog reaches production maturity with Views, hierarchical namespaces, and major performance improvements
  • Delta Sharing enables secure cross-organization data collaboration
  • Infrastructure improvements deliver zero-downtime pool management and proactive disk monitoring
  • Performance optimizations improve throughput and reduce latency across operations

For upgrade assistance, contact MinIO support through SUBNET.

0