All Products
Search
Document Center

Hologres:Guide to using trial instances in production environments

Last Updated:Feb 04, 2026

If you have used a Hologres trial instance and want to run more intensive performance tests or use the instance in a production environment, you must upgrade it to meet higher performance requirements. If you do not upgrade the instance, resource shortages during performance testing or production workloads can cause increased latency, slower processing, or service interruptions. This topic uses an 8-Core trial instance as an example to explain how to prepare a trial instance for a production environment.

Step 1: Upgrade the instance

Hologres 8-Core instances are intended only for evaluation and functional testing. Do not use them for performance testing or production workloads.

To run performance tests or deploy production jobs, you must first upgrade the instance, for example, to a 64-Core instance. For more information, see Upgrade or downgrade instances.

Step 2: Modify the shard count

Hologres is a distributed, real-time data warehouse engine. Data is stored across multiple data partitions, known as shards, in the underlying storage system. A proper shard count allows for better utilization of distributed computing resources and supports parallel processing, which improves query performance.

An 8-Core Hologres instance has a default shard count of 2. After you upgrade the instance, the shard count remains 2 even though the number of compute nodes (Workers) increases. This configuration prevents the system from fully leveraging its distributed query capabilities. After the upgrade, you must adjust the shard count. For more information, see Manage table groups and shard counts.

Note

You can also perform the following steps in the HoloWeb console. For more information, see Manage table groups.

  1. Create a new Table Group with an appropriate shard count.

    For a 64-Core Hologres instance, a shard count of 40 is recommended. Run the following command to create a new Table Group and set its shard count:

    CALL HG_CREATE_TABLE_GROUP ('<new_tg_name>', <shard_count>);

    In this command, new_tg_name specifies the name of the new Table Group and shard_count specifies the number of shards.

  2. Set the new Table Group as the default group.

    Run the following command to set the new Table Group as the default. New tables that you create will automatically belong to this group.

    CALL HG_UPDATE_DATABASE_PROPERTY ('default_table_group', '<new_tg_name>');
  3. Migrate existing tables to the new Table Group.

    Existing tables still belong to the original Table Group. Run the following command to move an existing table to the new Table Group.

    -- Syntax for V1.1 and later
    CALL HG_MOVE_TABLE_TO_TABLE_GROUP('<table_name>','<new_tg_name>');
    
    -- Syntax for V0.10
    CALL HG_UPDATE_TABLE_SHARD_COUNT('<table_name>','<new_tg_name>');

(Optional) Resource isolation and high availability

If your workload requires read/write splitting or read-read separation, you can use the Hologres shared-storage multi-instance high availability (HA) solution. For more information, see Deploy primary and secondary instances for read/write splitting (shared storage).

For example, for a 64-Core instance, you can configure a 32-Core primary instance to handle data writes and create a 32-Core read-only secondary instance to serve query requests. This setup achieves read/write splitting. If multiple applications require queries, you can use additional read-only secondary instances to achieve read-read separation.

Next steps

After your jobs are running stably in the production environment, consider the following steps:

  • To save costs on compute resources, you can switch the billing method from pay-as-you-go to subscription. For more information about pricing differences, see Billing overview. To learn how to change the billing method, see Switch between billing methods.

  • To reduce storage costs, you can purchase Standard storage resources on a subscription basis. To do this, you must first switch the instance to the subscription billing method and then upgrade it. .

References