Go to the ES purchase page, select Deployment Mode > Multi-AZ, and set up a multi-AZ network.
- To enable multi-AZ disaster recovery, a cluster must first have at least three dedicated master nodes enabled; therefore, there must be at least three regions selected where multi-AZ disaster recovery is supported. Currently, this feature is only supported in certain major regions such as Beijing, Shanghai, and Guangzhou, and it will be gradually rolled out in other regions as new Tencent Cloud data centers are constructed.
- Other parameter settings are generally the same as those for single-AZ deployment.
Take Shanghai as an example. Select Multi-AZ as the deployment mode. Currently, only two AZs are supported, so you need to select two pairs of AZ and subnet.
The Node Qty will be automatically equal to a multiple of the number of AZs. To ensure the cluster stability and reliability, Dedicated Master Node is enabled by default. You can configure three or five dedicated master nodes and evenly distribute them in three AZs to ensure that the cluster can have optional primary nodes more than the required number.
To ensure that multi-AZ disaster recovery takes effect, you need to follow the principles below:
ES automatically deploys the purchased data nodes evenly among the selected AZs, and the deployed data nodes can perceive and identify the AZs. This feature distributes the replicas of your data across multiple AZs so as to make sure that there is only one replica in a single AZ.
ES offers load balancing within VPC, which allows you to connect to the cluster through the provided VIP for reading and writing data and controlling the cluster through ES APIs.
This makes sure that when a node is down, or an AZ of a data center is not available, the problematic node will be automatically removed, so that it will not be requested by the client. This helps implement imperceptible failover in case of an AZ failure, thus improving the stability of your business.
In order to improve the reliability of your cluster, you must create at least three dedicated master nodes when using multi-AZ disaster recovery, and they must be distributed in three different AZs. Even if you select two AZs to deploy data nodes, an additional one will be automatically chosen to deploy a dedicated master node. This deployment scheme helps ensure that when an AZ becomes unavailable, your cluster still has electable nodes which form a quorum (majority) for the election of a master node.