メインコンテンツまでスキップ
バージョン: User Guides (Cloud)

April 2026 Release Notes

2026-04-11

Global Cluster​

Global Cluster now fully supports region-level disaster recovery failover with refined platform capabilities.

  • Failover: You can now trigger a failover at any time when the primary cluster becomes unavailable. After failover, new secondaries are automatically recreated.

  • Independent Replica Scaling: Primary and secondary clusters can now independently manage replica counts, Dynamic Scaling, and Schedule Scaling configurations.

  • Easy Conversion: You can now seamlessly convert between a Global Cluster and a regular dedicated cluster.

  • Audit Logging: All Global Cluster topology changes — including creation, switchover, failover, and secondary management — are now recorded in the audit log. For details, refer to Global Cluster Explained and Switchover and Failover.

Collection-Level Metrics​

The following metrics now support collection-level breakdown, helping you pinpoint performance issues and plan capacity for individual collections:

  • QPS (Read/Write)

  • Latency (Read/Write, Average and P99)

  • Entity Count

  • Loaded Entities

You can access collection-level metrics from the Console UI, Prometheus endpoint, or RESTful API.For details, refer to Metrics Reference and Integrate with Prometheus.

Access Logs
Public Preview
​

Zilliz Cloud now supports Access Logs for capturing query-level activity (Search, Hybrid Search, Query) on your clusters, designed for performance analysis and business insights. Key capabilities:

  • Configurable sampling — balance accuracy vs. storage cost (e.g., 1% sampling rate).

  • Customizable output fields — control log verbosity per entry.

  • Hot data identification — analyze returned primary keys to find frequently accessed records.

  • Structured JSON Lines format — ready for any analytics pipeline.For details, refer to Access Logs.

Maintenance Window​

The maintenance window experience has been redesigned with extended 4-hour minimum duration, email and in-console notifications at 7/3/1 days before upgrades, the ability to defer upgrades by 7 days, and availability for all Business Critical and Enterprise plans.For details, refer to Configure Maintenance Windows.

Cluster Admin Role​

A new Cluster Admin role lets you grant team members operational access to specific clusters without full project-level admin privileges.

  • Cluster Operations: Cluster Admins can perform day-to-day operational tasks — including scaling, suspend/resume, backup/restore, and DB user management.

  • Per-Cluster Scoping: The role can be assigned to specific clusters, enabling fine-grained separation of duties across environments and workloads.

  • Note: Customized API Keys do not currently support the Cluster Admin role. For details, refer to Manage Project Users.

Zilliz Cloud BYOC supports Tiered-Storage Cluster​

You can now create Tiered Storage clusters in BYOC deployments. To support this, BYOC projects introduce Tiered Query Node group configuration, allowing you to independently set instance types, node counts, and scaling for Query Nodes of Tiered Storage clusters. For details, refer to Deploy BYOC on AWS.

Enhancements​

  • You can now sort collection data previews in ascending or descending order by primary key, numeric, or other scalar fields. For details, refer to the Preview collection data section in Manage Collections (Console).