# MinIO AIStor RELEASE.2026-02-02T23-40-11Z

Released: February 2, 2026

This release represents a major milestone for MinIO AIStor's lakehouse capabilities with the S3 Tables/Iceberg Catalog and a complete implementation of the Delta Sharing protocol. The Iceberg Catalog now supports Views API, multi-level namespaces, and includes significant performance improvements through msgpack storage format, namespace sharding, and optimized locking. Delta Sharing brings secure cross-organization data sharing with full protocol v1.0 compliance including Databricks compatibility. Infrastructure improvements include pool hot-reload, S.M.A.R.T. disk monitoring, coalesced distributed locks, and async LDAP/OIDC initialization for improved Kubernetes deployments.

---

## Downloads

### Binary Downloads

| Platform | Architecture | Download |
|----------|--------------|----------|
| Linux | amd64 | [minio](https://dl.min.io/aistor/minio/release/linux-amd64/minio) |
| Linux | arm64 | [minio](https://dl.min.io/aistor/minio/release/linux-arm64/minio) |
| macOS | arm64 | [minio](https://dl.min.io/aistor/minio/release/darwin-arm64/minio) |
| macOS | amd64 | [minio](https://dl.min.io/aistor/minio/release/darwin-amd64/minio) |
| Windows | amd64 | [minio.exe](https://dl.min.io/aistor/minio/release/windows-amd64/minio.exe) |

### Package Downloads

| Format | Architecture | Download |
|--------|--------------|----------|
| DEB | amd64 | [minio_20260202234011.0.0_amd64.deb](https://dl.min.io/aistor/minio/release/linux-amd64/minio_20260202234011.0.0_amd64.deb) |
| DEB | arm64 | [minio_20260202234011.0.0_arm64.deb](https://dl.min.io/aistor/minio/release/linux-arm64/minio_20260202234011.0.0_arm64.deb) |
| RPM | amd64 | [minio-20260202234011.0.0-1.x86_64.rpm](https://dl.min.io/aistor/minio/release/linux-amd64/minio-20260202234011.0.0-1.x86_64.rpm) |
| RPM | arm64 | [minio-20260202234011.0.0-1.aarch64.rpm](https://dl.min.io/aistor/minio/release/linux-arm64/minio-20260202234011.0.0-1.aarch64.rpm) |

### FIPS Binaries

| Platform | Architecture | Download |
|----------|--------------|----------|
| Linux | amd64 | [minio.fips](https://dl.min.io/aistor/minio/release/linux-amd64/minio.fips) |
| Linux | arm64 | [minio.fips](https://dl.min.io/aistor/minio/release/linux-arm64/minio.fips) |

### Container Images

```bash
# Standard
docker pull quay.io/minio/aistor/minio:RELEASE.2026-02-02T23-40-11Z
podman pull quay.io/minio/aistor/minio:RELEASE.2026-02-02T23-40-11Z

# FIPS
docker pull quay.io/minio/aistor/minio:RELEASE.2026-02-02T23-40-11Z.fips
podman pull quay.io/minio/aistor/minio:RELEASE.2026-02-02T23-40-11Z.fips
```

### Homebrew (macOS)

```bash
brew install minio/aistor/minio
```

---

## S3 Tables / Iceberg Catalog

### Iceberg Views API

Complete implementation of the Iceberg Catalog Views API per the official specification (#2060, #2276, #2813, #2725):

**Core Operations:**

- **CreateView**: Create new views with SQL representation and schema
- **LoadView**: Retrieve view metadata including SQL, schema, and properties
- **ReplaceView**: Update existing view definition
- **DropView**: Delete view with optional purge
- **CommitView**: Atomic view updates with conflict detection (#2276)
- **RegisterView**: Register external views as AIStor extension (#2725)

**Implementation Details:**

- View metadata stored in Iceberg metadata format under `metadata/` prefix
- Views support multiple SQL representations for different query engines
- CommitView returns HTTP 409 on concurrent modification conflicts (#2882)
- Views share namespace with tables but maintain separate registry
- Full integration with TableAuthSys for authorization (#2280)
- TraceTables tracing for view operations (#2407)

### Multi-Level Namespaces

Hierarchical namespace support enabling complex organizational structures (#2491):

**Features:**

- Nested namespace paths (e.g., `catalog.database.schema`)
- Reserved namespaces for system use (#2385)
- ListNamespace policy restricted by parent namespace (#2812)
- HEAD namespace endpoint for existence checks (#1987)
- Namespace properties update functionality (#2278)
- Parent namespace validation in ListNamespaces (#2380)

**Implementation:**

- Multi-namespace character escape handling in SigV4 signatures (#1679)
- Dot notation for IAM policy condition matching (#2730)
- Namespace registry sharding to avoid contention (#2609)
- Synchronous cleanup on DeleteNamespace to prevent orphaned data (#2827)

### Performance Optimizations

#### Msgpack Storage Format

Replaced JSON with msgpack for all on-disk table/view metadata (#2703):

- **7,670+ lines** of generated serialization code
- Faster serialization/deserialization
- Smaller storage footprint
- Backward-compatible migration on first access

#### Namespace Sharding

Sharded namespace.json into 16 separate files to eliminate contention (#2609):

- Prevents lock contention on large warehouses
- Reduces file size for warehouses with many tables
- Uses xxhash for consistent shard distribution
- Parallel shard loading for ListTables

#### Shard-Based Pagination

New pagination implementation for ListTables and ListViews (#2889):

- Load single shard per page instead of all 16 shards
- Binary search for pagination token lookup (#2804)
- Cross-node token routing via NodeName encoding
- 15-minute cache TTL with 3-minute idle expiry
- Memory reduction: 16x for first page requests

#### Metadata Caching

- **Metadata pointer caching** for load-table operations (#2900)
- **Registry file read optimization**: Read once instead of multiple times (#2646)
- **MakeTableLive consolidation**: Single shard load for multiple operations (#2775)
- **Redundant metadata lookup elimination** in hot paths (#2774)
- **Memory allocation reduction** using pooled msgp readers (#2776)
- **Singleflight for registry shards**: Coalesce concurrent shard reads to reduce I/O amplification (#2912)
- **Single table lazy loading** for S3 operations (#2829)

#### Locking Improvements

- **Locking rework** to reduce number of locks and contention (#2789)
- **Per-warehouse namespace locks** replace global lock for RegisterTable/RegisterView (#2884)
- **Non-blocking locks** for multi-table transactions (#1771, #1748)
- **Catalog metadata max parity** for improved durability (#2771)

### Transaction Recovery

#### Unified Recovery Framework

Implemented Recoverable interface on Transaction for unified recovery (#2548):

- Single recovery code path for all transaction types
- Eliminates 200+ lines of duplicated recovery code
- Clear separation of concerns in transaction hierarchy (#1806)
- State helpers: LoadState, CheckForZombieTransaction

#### Atomic RenameTable

RenameTable now fully atomic with transaction log recovery (#2724):

- **3,975+ lines** of implementation and tests
- Two-phase commit with persistent transaction log
- Automatic recovery on server restart
- Handles partial failures and network partitions
- Prevents table name conflicts during rename

#### Transaction Hardening

- **Transaction logging hardening** (#2205)
- **Multi-table recovery rollback prevention** when participant dropped (#2636)
- **Validation before atomic log write** in multi-table transactions (#2625)
- **Orphaned staged-to-live log prevention** (#2641)
- **Staged table lifecycle enforcement** (#1700)

### Authentication and Authorization

#### TableAuthSys

New dedicated authentication system with caching for table operations (#2249, #2286):

- **1,438+ lines** of implementation
- Per-table/view authorization checks
- Cache with notifications for invalidation (#2286)
- Reduces repeated IAM policy evaluations
- Integrated with S3 standard SigV4 authentication (#1672)

#### IAM Enhancements

- **RegisterTable permission**: New s3tables IAM action (#2454)
- **Authorization filtering**: Tables and namespaces listing filtered by user permissions (#2203)
- **Multi-table transaction privilege checks**: Verify permissions on each table in transaction (#2352)
- **Auth filtering for Create\* operations** (#2519)
- **S3 Tables IAM policy resource handling** (#1717)
- **Proper s3tables policy actions** (#1675)
- **Explicit ListTables/ListViews permission** requirement (#2787)
- **TableMetrics authorization check** (#2359)

### Error Handling

Comprehensive error handling improvements for Iceberg REST Catalog compatibility (#2221):

| HTTP Code | Condition                              | PRs          |
| --------- | -------------------------------------- | ------------ |
| 409       | Concurrent modification conflicts      | #2834, #2882 |
| 404       | Namespace deleted during operation     | #2886        |
| 404       | Table/view not found (wrapped errors)  | #2824        |
| 400       | Invalid create-table/view requests     | #1847, #2487 |
| 400       | CommitMultiTableTransaction validation | #2487        |
| 400       | Warehouse bucket deletion via S3 API   | #2432, #2179 |

**Additional Fixes:**

- Response codes aligned with iceberg-go spec (#2792)
- View conflict detection in CommitTable for RCK compatibility (#2517)
- Error messages with context in S3 Tables/Iceberg catalog (#2363)
- Namespace validation error handling (#1667)
- Staged table cleanup during namespace deletion (#2399)

### Warehouse Bucket Management

Enhanced isolation and protection for warehouse buckets:

- **Filter from S3 and admin listings** (#2915)
- **Allow S3 list/head operations** for compatibility (#2881)
- **Prevent as replication targets** (#2717)
- **Block bucket ops** (versioning, lifecycle, encryption) on warehouse buckets (#2695)
- **Exclude from Replication and ListBucket** (#2658)
- **IsWarehouse field** in bucket metadata for detection (#2668)
- **Persist warehouse bucket tag** via globalBucketMetadataSys (#2644)
- **Validate bucket configs** before warehouse upgrade (#2757)
- **Notify peers** to load bucket metadata after warehouse creation (#2890)

### Inventory Integration

Support for S3 Inventory with AIStor Tables (#1867):

- **3,392+ lines** of implementation
- Generate inventory reports for Iceberg tables
- Iceberg metadata format output
- Batch catalog record count and file splitting fixes (#2452)
- createTablesAdapter for inventory Iceberg format support

### Additional Tables Features

- **ETag support** for table operations (#2507)
- **X-Minio-Tables-Include-Paths header** for ListTables/ListViews to include full object paths (#2885)
- **Namespace type refactoring** with proper type safety (#2780)
- **View creation with staged table**: Allow view creation when staged table exists with same name (#2393)
- **RegisterTable path overlap check**: Uses disk-based checks instead of in-memory for reliability (#2420)
- **Last consistent state on LoadTable**: Return last consistent state during recovery (#1911)
- **Iceberg metadata file naming**: Correct naming convention for metadata files (#2499)

### Client Compatibility

#### Spark

- **Per-bucket purge-on-delete** for Spark clients via versioning config (#2779)
- **DROP TABLE defaults purgeRequested to true** (#2848) - workaround for Spark bug where PURGE flag not forwarded
- **Distributed setup** for PySpark tests (#1720)

#### Trino

- **Trino CI/CD tests** (#1730)
- **Allow Trino to set table location** (#1723)
- **Trino notebook fixes** (#1837)

#### PyIceberg

- **Apache Iceberg RCK and PyIceberg catalog tests** (#2189)
- **Distributed setup** for PyIceberg tests (#1720)
- **iceberg-go fork updates** (#2777, #2466, #2022, #1661)

### Metrics and Observability

- **Tables Catalog v3 metrics** integration (#2126)
- **TTFB and request count metrics** for Iceberg APIs (#2614)
- **Tables/delta metrics fixes** (#2738)
- **Admin trace support** for AIStor tables (#2165)
- **TraceTables** instead of TraceObject in DeleteTable (#2407)

---

## Delta Sharing

### Protocol Implementation

Full implementation of Delta Sharing open protocol v1.0 for secure cross-organization data sharing (#2450):

**Scope: 19,740+ lines across 62 files**

#### Core Protocol Compliance

- Complete REST API per Delta Sharing protocol specification
- Bearer token authentication (RFC 6750)
- `delta-sharing-capabilities` header negotiation
- Response formats: parquet (default) and delta
- NDJSON streaming responses with proper Content-Type
- `Delta-Table-Version` header on all table responses
- Pagination (maxResults, pageToken, nextPageToken)
- Case-insensitive name matching per spec

#### Delta Lake Integration

- **Delta log parser** (1,000+ lines): Read Delta Lake transaction logs
- **Predicate pushdown** (1,605+ lines): Query optimization for efficient data retrieval
- **Snapshot cache**: Metadata caching for Delta tables
- **Uniform reader**: Read both Delta and Iceberg table formats

#### Authentication

- OAuth 2.0 client credentials support
- JWT token isolation for Delta Sharing API endpoints (#2743)
- Token management and credential regeneration (#2694, #2711)
- Refresh token support with secure storage

### Databricks Compatibility

- **endStreamAction inclusion** for streaming responses (#2583)
- **Proper UUID formatting** for table IDs
- **Consistent table IDs** for uniform tables

### Admin API

- **Share management**: Create, list, update, delete shares
- **Share name handling** improvements (#2497)
- **Multi-table share policy generation** fix for same-bucket tables (#2627)
- **Admin tracing** for Delta Sharing REST APIs

### Error Handling

- **Page token validation** (#2607)
- **Error code fixes** for protocol compliance (#2727)
- **Delta Sharing tags** in logs with reduced redundant error logging

### Metrics

- **Prometheus TTFB and request count metrics** (#2614)
- **Delta sharing Prometheus v3 metrics** (#2541)

### Documentation

Comprehensive documentation in `docs/aistor-delta-sharing/`:

- README.md (470 lines): Overview and concepts
- api-reference.md (336 lines): Complete API documentation
- client-setup.md (412 lines): Client configuration guides
- user-guide.md (361 lines): Usage examples and best practices

---

## Infrastructure Improvements

### Pool Hot-Reload

Dynamically add or remove storage pools without server restart (#2302):

- Update `config.yaml` and send `SIGHUP` to apply pool changes
- All nodes validate changes before applying
- Heal buckets to new pool automatically during hot reload (#2642)
- Extended `config.yaml` supports most server CLI flags (#2597)

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

Pure Go S.M.A.R.T. monitoring for NVMe and SATA drives (#2629):

- Metrics at `/minio/metrics/v3/system/drive/smart`
- Health status tracking (healthy, warning, critical)
- Failure risk prediction based on Backblaze empirical data
- Requires `CAP_SYS_ADMIN` and `CAP_DAC_OVERRIDE` capabilities or root

### Thread Pressure Monitoring

Proactive thread exhaustion protection (#2631):

- Health endpoints return HTTP 429 at critical pressure (≥85% of kernel limit)
- `X-Minio-Server-Status: thread-pressure` header for orchestrator detection
- Configurable thresholds via `mc admin config set`

### Coalesced Distributed Locks

Read lock coalescing for reduced RPC overhead (#2768):

- Multiple goroutines share single distributed RLock via reference counting
- Reduces N distributed RPCs to 1 for N concurrent readers
- Addresses ServerBusy errors under high concurrency

### Async LDAP/OIDC Initialization

Server starts when identity providers are temporarily unreachable (#2781):

- Configuration errors still block startup
- Connectivity errors allow startup with providers in pending state
- Prevents cascading container restarts in Kubernetes

### Extended Erasure Coding

Support for up to 32 drives per erasure set (#2376):

- Default remains 16 drives for optimal performance
- Configure via `MINIO_ERASURE_SET_DRIVE_COUNT=17-32`

### Performance Optimizations

- **DeleteVersions Grid RPC migration**: Migrated from HTTP REST to Grid RPC for improved performance (#2640)
- **Heap-based k-way merge for ListObjects**: O(n log k) replaces O(n\*k) linear scan when merging entries from multiple erasure sets (#2544)
- **MRF abandoned children healing**: Moved from inline scanner to background MRF with controlled concurrency (#2558)
- **MRF reading and parsing optimization** (#2654)
- **DeleteObject disk operations reduction**: Eliminates duplicate GetObjectInfo calls, reducing disk reads by factor of disks_per_set \* pools (#2438)
- **PUT request hot path optimization** for high performance networks (#2638)
- **Heap escape reduction** in hot paths (#2596)
- **Lock contention reduction**: List cache uses atomic operations and sync.Map; DiscoverTables rate-limited to max 50 concurrent goroutines (#2910)
- **Path.Clean allocation elimination**: Allocation-free path comparison in metacache merge (#2660)
- **Regex precompilation**: Table/namespace name validation regex moved to package-level, eliminating ~1200ns/op overhead per request (#2650)
- **PutObject temp directory sharding**: 16-way sharding of upload temp directories to reduce contention on `.minio.sys/tmp` (#2577)
- **24h runtime metrics**: Added runtime, memory, process, scanner, OS and network metrics for last 24 hours (#2705)

### Monitoring Enhancements

- **Per-peer internode metrics**: RTT, reconnect counts, queue length for identifying misbehaving nodes (#2610)
- **Silent data corruption metrics**: Prometheus metrics for detecting bit rot and corruption (#2514)
- **API version in healthinfo response** (#2897)

### Replication Enhancements

- **S3 RTC threshold events**: ObjectReplicationMissedThreshold and ObjectReplicationReplicatedAfterThreshold events per AWS S3 RTC spec (#2606)
- **Configurable threshold**: `MINIO_REPLICATION_EVENT_THRESHOLD` (default 15m)
- **Site replication UI**: Configure site replication from Console without mc (#2475)

### Additional Infrastructure

- **Pool filter for inspect-data**: Extract data from specific pool via `pool` query parameter (#2759)
- **Scanner refreshes bucket configs** every minute during long-running scans (#2808)
- **Avoid pools with offline drives** near EC threshold (#2647)
- **Drive max_timeout KVS config** honored (#2766)
- **Case-insensitive IAM map** for Windows and macOS (#2618)
- **Batch key rotation** for plaintext objects (#1843)
- **eBPF loader rewritten in Go** using cilium/ebpf (#2534)
- **Homebrew support**: macOS users can install via Homebrew formula (#2836)
- **Build info flag**: `--build-info` flag for extended build information (#2603)

---

## Experimental: Server-Side Logging

> **Note:** This feature is experimental. The backend implementation may change in future releases as we evaluate scale characteristics. This is an opt-in feature disabled by default.

### Persistent Log Recorder

Server-side persistent logging for API, Error, and Audit events (#2173, #2463):

**Internal Disk Recording:**

- `log_api_internal`: API request/response logging
- `log_error_internal`: Error event logging
- `log_audit_internal`: Audit trail logging

**External Log Targets:**

- Webhook targets: `log_api_webhook`, `log_error_webhook`, `log_audit_webhook`
- Kafka targets: `log_api_kafka`, `log_error_kafka`, `log_audit_kafka`
- Configurable batching for improved throughput
- Persistent queue recovery after restart

**Configuration:**

```bash
# Enable internal API logging
mc admin config set myminio log_api_internal enable=on

# Configure webhook target
mc admin config set myminio log_api_webhook:mytarget \
    endpoint=https://logs.example.com/api \
    auth_token=xxx
```

**Important Notes:**

- Disabled by default (#2411)
- Backend storage format subject to change
- Monitor disk usage when enabling internal logging
- Race condition and lock handling fixes included (#2752, #2751, #2851, #2917)

---

## Bug Fixes

### Security

- **CVE fixes**: Go 1.25.6 security updates for zip handling vulnerabilities (#2758)
- **Environment variable redaction**: Variables containing "passwd" redacted in logs (#2891)
- **Kubernetes nil pointer**: Fixed dereference when HTTP request fails (#2831)

### Race Conditions

- **WaitGroup race** in netperf (#2753)
- **Storage metric persistence** load race at startup (#2760)
- **Historic metrics initialization** race (#2805)
- **Lock file race** during drive rotation in log recorder (#2751)
- **Log recorder overflow** notification race (#2752)
- **Scheduler PutConfig** race recreating run metadata (#2739)
- **License management** race - health check before license validated (#2765)

### Replication

- **Status set correctly** in all cases (#2677)
- **Disabled target filtering** from backlog output (#2769)
- **Resync cancellation** validation relaxed (#2713)
- **EdgeSyncBeforeExpiry** blocking ILM expiry (#2619)
- **Inline checksum filter** for replication (#2648)

### Cluster Operations

- **Decommission metadata skip**: Skip all .minio.sys metadata in post-verification (#2906)
- **Decommission inode tracking** for progress monitoring (#2793)
- **Decommission object readability**: Verify object readability to handle inconsistent metadata (#2699)
- **Pool hot-reload S3PeerSys** update (#2635)
- **EC mismatch merging** fix (#2807)
- **Storage consistency** and error handling (#2590)

### Identity and Access

- **OpenID access key claims** deletion fix (#2823)
- **OpenID access key list/info** retrieval fix (#2620)
- **Group member normalization** fix (#2633)
- **JWKS pull fallback** client (#2671)
- **Skip root user service accounts** during site replication init (#2676)

### Licensing

- **License management consolidation** into single licenseManager (#2765, #2691)
- **License config order**: Set before token lookup (#2810)
- **Free license restriction** on distributed setups (#2643)

### Metrics

- **TTFB accuracy**: Excludes request read time (#2843)
- **Inventory sequence number** fix (#2750)
- **Missing metric descriptors** added (#2622)
- **Metrics initialization** before starting load (#2805)

### Startup and Configuration

- **TLS certificate error messages**: Improved actionable error messages for TLS misconfigurations in DEB/RPM deployments (#2904)
- **Server start without certs directory**: Server starts when certs directory is unavailable (#2632)
- **Version format panic**: Fixed panic when version is not set or has unknown format (#2600)

### ILM

- **Deleted tier handling**: Handle deleted tier during free-version cleanup (#2599)
- **Batch job start validation**: Fixed incorrect job start validation (#2381)

### RingBuffer

- **Flush deadlock fix**: Fixed Flush() deadlock when minWriteTo was set with partial buffer (#2709)

### Network

- **Netperf error handling**: Don't set error for individual peer connection failures (#2907)
- **Coalesced write lock error**: Correct error message on write lock contention (#2802)

---

## Important Notes

### Tables: DROP TABLE Behavior

DROP TABLE defaults `purgeRequested` to `true` (#2848):

- Workaround for Spark bug where PURGE flag not forwarded to REST catalog
- Matches behavior of Lakekeeper and Tabular.io catalogs
- Related to Apache Iceberg issue #11023

### Tables: On-Disk Format

Tables metadata uses msgpack storage format (#2703):

- Efficient serialization with smaller storage footprint
- Faster read/write operations

### Tables: HTTP Response Codes

Tables APIs return proper HTTP error codes per Iceberg REST Catalog spec:

- **409 Conflict**: Concurrent modification
- **404 Not Found**: Namespace/table/view not found
- **400 Bad Request**: Invalid requests

### v3 Metrics Bucket Limit

The v3 metrics wildcard endpoint now returns up to 1000 buckets per request (#2941), increased from the previous limit of 10. Use the `offset` query parameter for pagination when you have more than 1000 buckets.

---

## Upgrade Instructions

For detailed upgrade instructions: https://docs.min.io/enterprise/aistor-object-store/upgrade-aistor-server/

Platform-specific guides:

- **Linux/Bare Metal**: https://docs.min.io/enterprise/aistor-object-store/upgrade-aistor-server/upgrade-aistor-linux/
- **Kubernetes with Helm**: https://docs.min.io/enterprise/aistor-object-store/upgrade-aistor-server/upgrade-aistor-kubernetes-helm/

### New Configuration Options

**Environment Variables:**

- `MINIO_ERASURE_SET_DRIVE_COUNT` - Set EC set size to 17-32 drives
- `MINIO_REPLICATION_EVENT_THRESHOLD` - Configure S3 RTC threshold for replication events (default: 15m)

**Versioning Config:**

- `PurgeOnDelete: spark` - Enable purge-on-delete per bucket for Spark clients (#2779)

**Experimental Log Recording** (via `mc admin config`):

- `log_api_internal`, `log_error_internal`, `log_audit_internal`
- `log_api_webhook`, `log_error_webhook`, `log_audit_webhook`
- `log_api_kafka`, `log_error_kafka`, `log_audit_kafka`

### Implementation Notes

**Tables API Clients:**

- Handle HTTP 409, 404, 400 responses per Iceberg REST Catalog spec
- CommitView and CommitTable return 409 on concurrent modification conflicts

**Spark Users:**

- DROP TABLE purges by default (workaround for Spark PURGE flag issue)
- Per-bucket `PurgeOnDelete: spark` option available for granular control

**S.M.A.R.T. Monitoring:**

- Requires root or Linux capabilities: `cap_sys_admin,cap_dac_override+ep`

### Support

- SUBNET Support: https://subnet.min.io
- Documentation: https://docs.min.io