Cluster high availability

Updated at:
Copy as MD

ApsaraDB for SelectDB provides a high availability (HA) virtual cluster feature that protects against zone-level failures. A virtual cluster binds two physical clusters — a primary cluster and a standby cluster — in different zones. When the primary cluster becomes unavailable, the system automatically fails over to the standby cluster.

The HA architecture operates at two levels:

  • Storage layer: Zone-redundant storage replicates data across zones, ensuring zero data loss (Recovery Point Objective (RPO) = 0) when your client has a retry mechanism.

  • Compute layer: The virtual cluster monitors the primary cluster and automatically triggers failover within 10 to 20 seconds (Recovery Time Objective (RTO) ≤ 20 s).

Benefits

MetricValueCondition
RTO≤ 20 sAutomatic failover triggered on zone-level failure
RPO0Client must have a retry mechanism

Without an HA virtual cluster, recovery from a zone-level failure depends on restoring a single physical cluster, which can take significantly longer.

Prerequisites

Before you begin, ensure that you have:

  • Milvus version 4.0.7 or later

  • A zone-redundant (multi-zone) deployment architecture

    Note

    Zone-redundant storage is in public preview. To enable it for your account, submit a ticket to be added to the allowlist.

Create an HA virtual cluster

Before creating an HA virtual cluster, prepare a primary cluster and a standby cluster. Both clusters must be in the Running state and located in different zones. If you haven't created these clusters yet, see Create a cluster.

  1. Log on to the ApsaraDB for SelectDB console.

  2. In the upper-left corner, select the region where your instance is located.

  3. On the Instances page, click the Instance ID/Name of the target instance to go to the Instance Details page.

  4. In the left navigation pane, click Cluster Management.

  5. On the Cluster Management page, click Create Cluster.

  6. In the Create Cluster panel, click HA Virtual Cluster, and configure the following parameters.

    ParameterDescription
    Cluster NameThe name of the HA virtual cluster. The name must be unique within the instance, start with a letter (case-insensitive), and contain only letters, digits, and underscores (_). Consecutive underscores are not allowed. Maximum length: 64 characters. The name cannot be changed after creation.
    Primary ClusterThe physical cluster that serves traffic under normal conditions.
    Standby ClusterThe disaster recovery cluster that takes over if the primary cluster fails. Use the same specifications for the primary and standby clusters.
  7. Click OK.

Manage an HA virtual cluster

On the Cluster Management page, the following operations are available for an HA virtual cluster:

OperationDescription
Modify Cluster ConfigurationChange the primary and standby clusters bound to the virtual cluster.
DeleteRemove the virtual cluster. This operation cannot be undone.
Restart, Scale, or PauseApply to the primary and standby clusters within the virtual cluster.

Related API operations

What's next