Scheduled scaling (Beta) - Hologres - Alibaba Cloud Documentation Center

Hologres virtual warehouse instances support scheduled scaling. After you configure a scaling plan for a virtual warehouse, Hologres automatically scales out or in as scheduled. This helps improve instance stability, increase resource utilization, and reduce costs. This topic describes how to use scheduled scaling.

Use cases

For example, your virtual warehouse instance includes three virtual warehouses: init warehouse, warehouse 1, and warehouse 2. The data mid-end team uses the init warehouse to perform large-scale extract, transform, and load (ETL) operations, import a large amount of data offline, import data within minutes, or import data in real time. Business team A uses warehouse 1 to query data. Business team B uses warehouse 2 to query data. In this example, you can use scheduled scaling based on the following rules:

Data import within minutes and real-time data import: Use the init warehouse. If the amount of data to be imported throughout the day is stable, scheduled scaling is not required. If the amount of data that is imported during a fixed period of time (less than 16 hours) every day is significantly large, we recommend that you configure a scheduled scaling plan to provide elastic computing resources during peak hours.
Data queries: Different business teams use different follower virtual warehouses to perform queries. This ensures load isolation. If the numbers of query requests from the business teams are stable, scheduled scaling is not required. If the numbers of query requests per day periodically change, we recommend that you configure a scheduled scaling plan to provide elastic computing resources during peak hours.
Large-scale ETL operations and offline import of a large amount of data: We recommend that you use serverless computing resources. You can use serverless computing resources without the need to reserve computing resources. This helps improve the instance stability and lower the risk of out of memory (OOM) errors. You are charged only for the amount of serverless computing resources actually consumed. For more information about the Serverless Computing feature, see Serverless Computing. For more information about how to use the Serverless Computing feature, see Serverless Computing guide.

Terms

To learn more about instance resources and virtual warehouse resources, see Glossary.

Example:

Item	Resource details
Instance	Reserved resources: 96 CUs (64 CUs allocated + 32 CUs unallocated)
	Elastic resources: 32 CUs.
	Total computing resources: 96 + 32 = 128 CUs.
init_warehouse	Reserved resources: 32 CUs. Elastic resources: 16 CUs. Total computing resources: 48 CUs.
warehouse_2	Reserved resources: 32 CUs. Elastic resources: 16 CUs. Total computing resources: 48 CUs.

Pricing

Instance reserved resources: These are dedicated computing resources allocated to your instance. Charged based on your instance's billing method (subscription or pay-as-you-go).
Instance elastic resources: These are additional computing resources provisioned by the scheduled scaling feature. Billing formula: Cost = Actual elastic resources used (CU-hours) × Unit price. For unit price details, see Billing overview.
Note
- Fees are calculated and settled hourly. The system records elastic resource usage minute-by-minute, calculates the hourly total, generates a bill, and deducts the fees automatically.
- Elastic resources are separate from unallocated reserved resources. Even if your instance has unallocated reserved capacity, scheduled scaling will provision additional elastic resources rather than utilizing unallocated ones.

Limitations

Scheduled scaling is supported only by virtual warehouse instances.
Hologres V2.2.21 and later support scheduled scaling.

The following table describes the support for scheduled scaling in different regions.

Scheduled scaling is in public preview. If you're interested, submit a request.

Region	Support for scheduled scaling	Description
China (Hangzhou), China (Shanghai), China (Beijing), and China (Shenzhen)	Supported	You can directly use this feature after your application is approved.
China East 2 Finance, China North 2 Ali Gov 1, China South 1 Finance, Japan (Tokyo), Malaysia (Kuala Lumpur), and Indonesia (Jakarta)	Not supported	You cannot submit an application for using this feature.
China (Hong Kong), Singapore, Germany (Frankfurt), US (Silicon Valley), and US (Virginia)	Submitting a ticket to apply for using this feature	To use the feature, perform the following steps: Fill in the form to submit an application for using scheduled scaling. Submit a ticket.

Usage notes

To use scheduled scaling, you must use one of the following account types:
- Alibaba Cloud account or RAM users to which the AliyunHologresWarehouseFullAccess policy is attached. The AliyunHologresWarehouseFullAccess policy contains the read-only permissions on the Hologres console and the permissions to configure scheduled scaling. For more information about permission authorization, see Grant permissions on Hologres to a RAM user.
- Accounts with superuser permissions on the instance. For more information about permission authorization, see Grant the development permissions on a Hologres instance to a RAM user.
Scale-out and scale-in operations on virtual warehouses have certain impact on data queries and writes:
- In Hologres V2.2, when you use scheduled scaling to scale in or scale out a virtual warehouse as scheduled, data queries and writes on the virtual warehouse are interrupted for about 15 seconds.
- In Hologres V3.0 and later, when you use scheduled scaling to scale out a virtual warehouse as scheduled, data queries and writes on the virtual warehouse are not affected. The virtual warehouse is in the processing state. When you use scheduled scaling to scale in a virtual warehouse as scheduled, data queries and writes on the virtual warehouse are interrupted for about 15 seconds.
- In Hologres V3.1 and later, when you use scheduled scaling to scale in a virtual warehouse as scheduled, major data queries and writes on the virtual warehouse are not affected. However, if the scale-in is triggered while a large query or large write operation is running in the virtual warehouse, and the task needs to continue running for more than 60 minutes, the task will report an error at the 60th minute after the scale-in.
After you configure a scaling plan for a virtual warehouse, take note of the following items:
- You cannot suspend, delete, or scale in the virtual warehouse.
- You can scale out the virtual warehouse only in the Hologres console, not by manually executing the hg_alter_warehouse statement.
- You can create a virtual warehouse only in the Hologres console, not by manually executing the hg_create_warehouse statement.

Usage guide

Manage virtual warehouse resources

Go to the Virtual Warehouse Management page.
1. Log on to the Hologres console. In the top navigation bar, select a region from the drop-down list.
2. In the left navigation pane, choose Instances and click the ID of your instance.
3. In the secondary navigation menu on the left, click Virtual Warehouse Management.
View the instance resources.
On the Virtual Warehouse Resource Management tab, view the resource overview of the instance.
Manage resources.
- Click Create Virtual Warehouse. In the Create Virtual Warehouse dialog box, enter a name, set the resources to allocate, and click OK.
- Click Modify Configuration, Restart, Rebalance, or Stop for a running virtual warehouses. For more information about shard rebalancing, see Rebalance.
  Note
  The default virtual warehouse init_warehouse cannot be stopped.

Configure scaling plans

View scaling schedules

On the Virtual Warehouse Management page, select the Scheduled Scaling tab.
Select your instance or a virtual warehouse from the Instance/Virtual Warehouse drop-down list to view time series charts. This chart indicates scheduled scaling plans for an instance or virtual warehouse.

Manage a scaling plan

On the Scheduled Scaling tab, click on the left of the target virtual warehouse, and click Add Time Period. Configure the start time and end time of the effective scaling time period, set the amount of elastic computing resources, and then click Save in the Actions column.
Click Edit or Delete in the Actions column for an existing scaling plan.

Important

The amount of elastic computing resources cannot exceed the amount of reserved resources.
You can configure up to five scaling time periods for a virtual warehouse.
When you delete a scaling plan from a virtual warehouse, the virtual warehouse enters the processing state even if the resources of the virtual warehouse remain unchanged. In this process, data reads and writes on the virtual warehouse are not affected.

After you save a scaling plan for a virtual warehouse, the resources of the virtual warehouse are immediately scaled out or in if the current time falls in a related scaling time period. Examples:

Example 1:
Virtual Warehouse A has 64 CUs of reserved computing resources. You create and save a scaling plan for Virtual Warehouse A at 16:00.
In the scaling plan, 32 CUs of elastic computing resources are allocated, and the effective time period is from 15:00 to 18:00. In this case, the total amount of computing resources for Virtual Warehouse A is immediately scaled out to 96 CUs after you save the scaling plan.
Example 2:
Virtual Warehouse A has 64 CUs of reserved computing resources. You modify and save an existing scaling plan for Virtual Warehouse A at 16:00.
- Before the modification: 32 CUs of elastic computing resources are allocated, and the effective time period is from 15:00 to 18:00.
- After the modification: 32 CUs of elastic computing resources are allocated, and the effective time period is from 17:00 to 18:00.
After you save the modification, the total amount of computing resources for Virtual Warehouse A is immediately scaled in from 96 CUs to 64 CUs.

Monitor scheduled scaling

This section describes how to monitor the performance of scheduled scaling.

Execution logs of scaling plans

On the Virtual Warehouse Management page, click the Elastic Event Execution Log tab.
Select a time range. The execution information of scaling plans in the specified time range is displayed, including the execution time, virtual warehouse name, execution status, event type, amount of reserved computing resources, and amount of elastic computing resources.

Metrics

You can view the Warehouse_timed_elastic_cores(Count) metric in the Hologres console. This metric indicates the number of cores of elastic computing resources that are provided by an instance based on scheduled scaling. You can also configure alert rules for this metric. For more information, see Monitoring metrics.

CloudMonitor events

The scale-in and scale-out events triggered by Hologres time-based scaling plans are recorded in CloudMonitor.

Go to Event Center in the CloudMonitor console. On the System Event page, select Hologres from the product drop-down list in the Event Monitoring section. Then, you can view the time information of the following time-based scaling events:
- Instance:Warehouse:TimedElastic:Start: The schedued scaling of a virtual warehouse starts.
- Instance:Warehouse:TimedElastic:Finish: The schedued scaling of a virtual warehouse is finished.
- Instance:Warehouse:TimedElastic:Failed: The schedued scaling of a virtual warehouse fails.

Configure notifications and alert rules based on the events in Cloud Monitor. For more information, see Create a system event-triggered alert rule.

The following code shows the details of an Instance:Warehouse:TimedElastic:Failed event in CloudMonitor:

{
    "Status": "Failed",
    "InstanceName": "<instance_id>",
    "ResourceId": "<instance_resource_id>",
    "Content": {
        "ScaleType": "ScaleDown",
        "ScheduleId": "xxxxxx",
        "TimedElasticCPU": 0,
        "WarehouseId": "2",
        "WarehouseName": "<warehouse_name>"
    },
    "Product": "hologres",
    "Time": 1722852008000,
    "Level": "WARN",
    "RegionId": "<region>",
    "Id": "<event_id>",
    "GroupId": "0",
    "Name": "Instance:Warehouse:TimedElastic:Failed"
}

ActionTrail

Operations such as modifications on scaling plans in the Hologres console and scaling operations are logged in the ActionTrail console. For more information, see Event audit logs.