Convert general purpose instances to virtual warehouse instances-Hologres(Hologres)-阿里云帮助中心

Virtual warehouse instances offer finer-grained load isolation, more flexible permission management, and easier connection management than general-purpose instances. Each virtual warehouse has its own configuration, so you can tune resources independently per workload. This topic describes how to convert a Hologres general-purpose instance to a virtual warehouse instance.

Prerequisites and limitations

Before you start, confirm that your instance meets all of the following conditions:

Version: The instance runs Hologres V2.0.40 or later. If the version is earlier than V2.0.40, upgrade the instance manually in the Hologres console, or contact technical support by joining the Hologres DingTalk group. See Upgrade an instance and Obtain online support for Hologres.
Specifications: The instance has more than 8 vCPUs. Instances with 8 vCPUs cannot be converted. To proceed, change the instance specifications first. For more information, see View instance details and configure network.

Limitations:

Instance role: Read-only secondary instances cannot be converted directly.
Direction: Converting a virtual warehouse instance back to a general-purpose instance is not supported.

Conversion methods

Two methods are available. Choose based on whether your business can tolerate a brief write interruption.

Important

Run the conversion during off-peak hours. Service downtime during normal conversion is not covered by the SLA.

Item	Normal conversion	Hot conversion
Duration	Typically 10–20 minutes	Typically 10–30 minutes
Service availability	Fully unavailable	Read-only
Data writes via Realtime Compute for Apache Flink	Stop before conversion; restart after to avoid data loss. See Stop a job.
DataWorks / Blink jobs	No action needed. Failovers trigger automatically. Set failover retries to 10 or more. See Real-time synchronization node operations for DataWorks tasks and Hologres sink table for Blink jobs.
Endpoint	Unchanged, but the IP address may change
How to request	Submit a ticket or contact technical support. See How do I get more online support?

Convert a general-purpose instance

A general-purpose instance has a single endpoint. After conversion, the endpoint and all metadata — including user permissions — remain unchanged.

To start the conversion, submit a ticket or join the Hologres DingTalk group with the instance ID and your preferred operation time. Hologres O&M engineers will perform the conversion in the background.

After conversion, set up load isolation for your virtual warehouses. See (Recommended) Set up load isolation.

Convert a primary/secondary instance pair

A primary/secondary instance pair uses a multi-endpoint load isolation architecture. Virtual warehouse instances use a single-endpoint architecture. Before converting, you must consolidate traffic to the primary instance and release the secondary instances.

Assume an instance pair consists of a primary instance with 64 compute units (CUs) and two read-only secondary instances each with 32 CUs each. Conversion procedure:

Scale up the primary instance to 128 CUs — the combined total of the primary and secondary instances. This ensures the primary instance can absorb all traffic while you migrate workloads, without service interruption.
Migrate query traffic from the secondary instances to the primary instance.
Disassociate or release the secondary instances. For more information, see Instances page overview. The primary instance is now a standalone general-purpose instance.
Request a conversion, as in Convert a general-purpose instance.

After conversion, set up load isolation for your virtual warehouses. See (Recommended) Set up load isolation.

Parameter changes after conversion

Parameter	Before	After
Instance Type	General-purpose	Virtual warehouse
Number of Gateway Nodes	N/A	Default: CUs ÷ 32 (valid range: 2–8). Adjustable after conversion — see Manage virtual warehouses.
Computing Resources	Supported	Unchanged. All CUs are allocated to the default `init_warehouse` virtual warehouse.

Next steps

(Important) Configure alert rules

After conversion, your instance moves from the Hologres standard instance tab to the Hologres warehouse instance tab in the Cloud Monitor console (Cloud Resource Monitoring > Cloud Service Monitoring > Hologres). This update takes approximately 10 minutes.

If no alert rules exist, configure them for key metrics on the Hologres warehouse instance tab. See the Configure alert rules section.
If alert rules already exist, reconfigure them on the Hologres warehouse instance tab. Consider adding gateway-related metrics that are now available for virtual warehouse instances.

(Recommended) Set up load isolation

After conversion, all computing resources are assigned to the default init_warehouse virtual warehouse. To isolate different workloads — for example, separating online analytical processing (OLAP) queries from online serving — create dedicated virtual warehouses and route traffic accordingly.

The following example uses a 128-CU instance:

After the O&M engineers complete the conversion, all 128 CUs are allocated to init_warehouse. Scale up the instance first so that init_warehouse has enough capacity to absorb traffic while you build out the new virtual warehouses — this prevents service interruption during the transition. For example, increase the reserved computing resources to 192 CUs. The 64 CUs of unallocated capacity will be assigned to new virtual warehouses in the next step.
Create virtual warehouses for each workload. For example, create warehouse_olap with 32 CUs for OLAP and warehouse_serving with 32 CUs for online serving.
Grant the appropriate users access to each virtual warehouse. Then grant permissions on the table groups loaded to each warehouse. Switch the corresponding business traffic to the new virtual warehouses.
Scale down init_warehouse now that the other virtual warehouses are handling their respective workloads. For example, reduce init_warehouse to 64 CUs, then decrease the instance's reserved computing resources from 192 CUs to 128 CUs. All virtual warehouses remain available during the scale-down.

(Optional) Enable additional features

Enable the following features per virtual warehouse based on your requirements:

Scheduled scaling (beta): Automatically scale a virtual warehouse at scheduled times. See Scheduled scaling (beta).
Serverless Computing: Run on-demand compute without pre-provisioned resources. See Serverless computing.

References

上一篇: Multi-cluster and auto scaling (beta) 下一篇: Hologres Serverless