OpenSearch:Upgrade or downgrade

Configuration change notes

Configurable item	Description
Number of query nodes	You can increase or decrease the number of query nodes. If you want to decrease the number of query nodes, make sure that the new number is greater than or equal to the number of query nodes in use. (The console validates this and displays a prompt.) Example: If the current instance has 10 query nodes and 4 of them are in use, you can decrease the number of query nodes to a minimum of 4.
Query node specifications	Currently, only scale-up of query nodes is supported. This means that the number of cores and the memory size of the query nodes after the change must both be greater than or equal to those of the original query nodes. Important When the number of query nodes changes from 0 to a value greater than 0, OpenSearch currently supports only a disruptive change, and the service is unavailable during the change. Proceed with caution. After the change is complete, the status of the query nodes and data nodes in the console changes to 100% In Service, and the service becomes available again.
Number of data nodes	You can increase or decrease the number of data nodes. If you want to decrease the number of data nodes, make sure that the new number is greater than or equal to the number of data nodes in use. (The console validates this and displays a prompt.) Example: If the current instance has 10 data nodes and 6 of them are in use, you can decrease the number of data nodes to a minimum of 6.
Data node specifications	Currently, only scale-up of data nodes is supported. This means that the number of cores and the memory size of the data nodes after the change must both be greater than or equal to those of the original data nodes.
Availability edition	You can switch between the standard edition and the high-availability edition. Standard edition: deployed in a single zone. Zone-level disaster recovery is not supported. High-availability edition: deployed across multiple zones with two clusters that back each other up, providing higher availability. If a zone fails, traffic can be switched to another zone.

How table data is stored

Tables share data node resources (CPU, memory, and disk) for data storage and queries:

The preceding figure shows two vector tables. Table 1 has 4 shards, and Table 2 has 1 shard. When data is written to Table 1, the system evenly writes the data to the 4 shards based on the generated index. When data is written to Table 2, because it has only 1 shard, Table 2 becomes a broadcast table, and a complete copy of its index data is stored on each data node.

Precautions

• Upgrade and downgrade operations take effect immediately, and you are billed based on the new specifications. The service becomes available again when the service status of the query nodes and data nodes on the instance details page reaches 100%.

• Each change can only scale out or scale in. You cannot scale out some items while scaling in other items in the same change.

  Example: If the original instance has 2 query nodes and 2 data nodes:

  Changing to 1 query node and 3 data nodes is not supported.

  Changing to 1 query node and 1 data node is supported.

  Changing to 1 query node and 2 data nodes is supported.

  Changing to 3 query nodes and 3 data nodes, among other options, is supported.

• After you change the number of data nodes, you must modify the number of data node replicas or edit tables to adjust the number of shards on the Instance Details page to allocate resources.

• Number of query nodes in a single cluster = Number of query nodes × Number of replicas. Number of data nodes in a single cluster = Number of replicas × Number of data shards. When you scale out, scale out in multiples of the number of shards. (A data shard is the smallest unit block of data.)

• The instance is unavailable while it is being upgraded from a single node to multiple nodes. Make sure that the instance receives no production query traffic during the upgrade.

• If you change from the high-availability edition to the standard edition, the second cluster of the instance is deleted by default. If the configurations of the clusters in the two zones are inconsistent, changing to a single zone may affect the service. Proceed with caution.

• After the configuration change, the number of data nodes in a single zone must be greater than or equal to the number of data nodes currently in use in a single zone.

Procedure

1. On the Instances page, find the application in the application list, and choose Upgrade/Downgrade in the Actions column on the right:

2. On the configuration change page, select the availability edition, the number and specifications of query nodes, and the number and specifications of data nodes that you want to scale out or scale in, select Service Agreement, and then click Buy Now:

   The configuration change page also includes configuration items such as Data Node Instance Family (valid values: Cloud Disk, Local SSD, and GPU), Query Node Instance Family, Free Storage per Data Node, and Total Storage per Data Node.

3. After the configuration change is complete and the status of the query nodes and data nodes in the console changes to 100% In Service, the service becomes available again.