ApsaraDB for SelectDB supports scheduled elastic scaling and manual scaling for clusters. If your business has predictable peak and off-peak hours, you can use scheduled scaling to automatically adjust resources. This improves resource utilization and system performance. If the current cluster configuration does not meet your needs, you can also scale the cluster manually.
Overview of cluster scaling
Scaling type | Scenarios | Impact on cluster |
Scheduled elastic scaling | The business has predictable peak and off-peak hours. | During a scale-in, the cache space is scaled in proportionally with the compute resources. Cached data that exceeds the target cache space is evicted. This may cause response time jitter for some requests. |
Manual scaling | The current cluster configuration cannot meet business needs. | When scaling in the cache space, cached data that exceeds the target cache space is evicted. This may cause response time jitter for some requests. |
Prerequisites
The instance status is Running.
The status of the target cluster is Running.
There are no unpaid orders on the Alibaba Cloud account.
Precautions
If your Alibaba Cloud account has an overdue payment or an insufficient balance, scheduled scaling rules will not be executed, and manual scaling attempts will fail.
Read and write operations might be briefly unavailable during scaling. Scale your cluster during off-peak hours.
Note the following for scheduled elastic scaling:
Only pay-as-you-go clusters support scheduled elastic scaling.
Scheduled scaling rules are executed only when both the instance and the cluster are in the Running state. If an instance or cluster is in another state, such as Paused, Restarting, or Upgrading, the system attempts to retry the execution. If these retries fail for more than 30 minutes, the rule is not executed.
Billing
The fees described in this topic are for reference only. The actual fees are specified on your bill.
Each instance can contain one or more Backend (BE) clusters. A pay-as-you-go instance contains only pay-as-you-go clusters. A subscription instance can contain both subscription and pay-as-you-go clusters. Therefore, the billing for changing the configuration of a cluster in a subscription instance is different from that in a pay-as-you-go instance.
Change the configuration of a cluster in a subscription instance
Change type | Cluster billing method | Billing details |
Cluster scale-out | Subscription | The fee for scaling out a cluster = (Daily price after scale-out - Daily price before scale-out) × Number of remaining days until the instance expires. For more information about pricing, see Billing items and pricing. Note The number of remaining days is not an integer and is accurate to 12 decimal places. For example, if the remaining time is 31 days and 10 hours, it is recorded as 31.416666666667 days. |
Pay-as-you-go | After a pay-as-you-go cluster is scaled out, you are charged per hour based on the new configuration. For the hour in which the change occurs, the system bills you by the minute in segments. The bill is generated and payment is deducted at the end of that hour. For more information about pricing, see Billing items and pricing. | |
Cluster scale-in | Subscription | After you scale in a subscription cluster, Alibaba Cloud refunds the amount for the remaining subscription duration to your original payment method. |
Pay-as-you-go | After a pay-as-you-go cluster is scaled in, you are charged per hour based on the new configuration. For the hour in which the change occurs, the system bills you by the minute in segments. The bill is generated and payment is deducted at the end of that hour. For more information about pricing, see Billing items and pricing. |
Change the configuration of a cluster in a pay-as-you-go instance
Pay-as-you-go instances use a pay-as-you-go billing method. After you change the configuration of a cluster in a pay-as-you-go instance, the billing rule remains the same, and you are still charged per hour. For the hour in which the change occurs, you are billed by the minute. The bill is generated and payment is deducted at the end of that hour. For more information about pricing, see Billing items and pricing.
Scaling limits
Compute resource scaling
When you adjust compute resources, the cache space is automatically adjusted proportionally based on the current compute-to-cache ratio of the cluster. For example, for a cluster with 8 CPU cores and 200 GB of cache, if you scale out the compute resources to 16 CPU cores, the cache space is automatically adjusted to 400 GB. If you scale in the compute resources to 4 CPU cores, the cache space is automatically adjusted to 100 GB.
You cannot adjust compute resources independently while keeping the cache space unchanged.
Cache space scaling
Scale-out: You can scale out the cache space independently without changing the compute resources. You can also scale out the cache space proportionally when you scale out the compute resources.
Scale-in: You cannot scale in the cache space independently. To scale in the cache, you must scale in the compute resources proportionally at the same time.
To scale in only compute resources or only cache space, you can create a new cluster with the target specifications and then delete the old cluster.
Procedure
Each scaling operation takes about 10 minutes. You can refresh the page to view the cluster status. When the cluster status changes from Changing Configuration to Running, the scaling is complete.
Scheduled elastic scaling
Log on to the ApsaraDB for SelectDB console.
In the upper-left corner of the page, select the region where the instance resides.
On the Instance List page, click the ID of the target instance to go to the Instance Details page.
On the Instance Details page, click Cluster Management in the navigation pane on the left.
On the Cluster Management page, click Scale in the Actions column of the target cluster.
In the cluster scaling panel, click the Scheduled Elastic Scaling tab.
Follow the prompts in the panel to perform the following operations:
ImportantBefore you create or modify a scheduled elastic scaling rule, note the following limits:
The execution time of a rule cannot overlap with that of an existing rule.
The interval between rules must be at least 1 hour. Therefore, you can configure a maximum of 23 rules.
Adjacent rules cannot have the same target compute resource specifications. The last rule and the first rule in the rule list are also considered adjacent.
Enable or disable the scheduled elastic policy.
Click the Scheduled Elastic Policy switch in the upper-left corner of the panel.
NoteYou must have at least two rules to enable the scheduled elastic policy.
Enable: The system executes the created rules in sequence. In this state, manual scaling is not supported.
Disable: The created rules are not executed and are not automatically deleted.
Add a scheduled scaling rule.
ImportantAfter you add a rule, the system will not execute the ruleset if the scheduled elastic policy is disabled.
Click Add on the right side of the panel, configure the Execution Time and Target Compute Resource, and then click Save.
If you have not added any rules, you can also click Add Rule at the bottom of the panel to configure the parameters.
Modify a scheduled scaling rule.
Click
in the Actions column of the target rule, modify the Execution Time and Target Compute Resource, and then click Save.View scheduled scaling rules.
The scaling panel displays the list of scheduled elastic scaling rules by default. If you are on the Manual Scaling tab, click the Scheduled Elastic Scaling tab.
Delete a scheduled scaling rule.
Click
in the Actions column of the target rule, carefully read the content in the Are you sure you want to delete this rule? dialog box, and then click OK.After deletion, if the number of rules is less than two, the system automatically disables the scheduled elastic policy and prevents you from enabling it.
NoteIf the two rules adjacent to the target rule have the same target compute resource specifications, you cannot delete the target rule.
The last rule and the first rule in the rule list are also considered adjacent.
Parameter
Description
Rule number
A temporary number for the rule. By default, rules are numbered by execution time. If you add or modify a rule and its execution time falls between other rules, the system automatically re-sorts the entire rule list.
Execution Period
The period for executing the rule. The default value is Daily. You cannot change this value.
Execution Time
The time when the system schedules the rule for execution.
Target Compute Resource
The target compute resource specifications for the cluster to scale to, including CPU and memory.
CPU cores:
[4 cores, 1024 cores]Memory: The default value is 4 times the number of CPU cores, in GB.
Important1 CCU = 1 core and 4 GB of memory.
To request a higher quota, contact Alibaba Cloud technical support.
Manual scaling
Manual scaling is not supported when the scheduled elastic policy is enabled. To perform a manual scaling operation, you must first disable the scheduled elastic policy. After the manual scaling is complete, you must re-enable the policy.
Log on to the ApsaraDB for SelectDB console.
In the upper-left corner of the page, select the region where the instance resides.
On the Instance List page, click the ID of the target instance to go to the Instance Details page.
On the Instance Details page, click Cluster Management in the navigation pane on the left.
On the Cluster Management page, click Scale in the Actions column of the target cluster.
In the cluster scaling panel, click the Manual Scaling tab.
Configure the target specifications.
Parameter
Description
Target Compute Resource
The compute resources for a single cluster, including CPU and memory.
CPU cores:
[4 cores, 1024 cores]Memory: The default value is 4 times the number of CPU cores, in GB.
Important1 CCU = 1 core and 4 GB of memory.
To request a higher quota, contact Alibaba Cloud technical support.
Target Cache Space (GB)
The cache space for a single cluster.
Value range:
Minimum:
For target compute resources with 4, 8, 16, or 32 CPU cores, the minimum cache space is 100 GB.
For target compute resources with 64 or more CPU cores, the minimum cache space is
100 GB × Number of selected CPU cores / 32.
Maximum:
500 × Number of target compute resource cores
Unit: GB
Terms of Service
Read and check the Terms of Service.
Click OK.
Related API operations
Create a scheduled scaling rule
Modify a scheduled scaling rule
Delete a scheduled scaling rule