Upgrade instance configurations

Updated at:

Upgrade the edition, traffic, disk capacity, partitions, or network type of an ApsaraMQ for Kafka instance in the console.

Prerequisites

Make sure that:

  • The instance is in the Running or Not Deployed state

  • No traffic rebalancing tasks are in progress on the instance

Upgrade scenarios and risks

Scenario

Risk

Increase the disk capacity of a non-serverless instance

Brokers in the cluster restart one by one.

Upgrade the traffic specification of a non-serverless instance

Brokers in the cluster restart one by one.

Upgrade from Standard Edition to Professional Edition (High Write) or Professional Edition (High Read)

Brokers in the cluster restart one by one.

Add a public endpoint to a VPC-connected instance

Brokers in the cluster restart one by one.

Change the network type from Internet and VPC to VPC

Brokers in the cluster restart one by one.

Change the public bandwidth of a non-serverless instance

Brokers in the cluster restart one by one.

Change the partition or topic count of a non-serverless instance. Instances purchased on or after August 26, 2022 support partition changes only. Instances purchased before August 26, 2022 support topic changes only.

None.

Downgrade the minimum usage specification of a serverless instance

Auto scaling is suspended after the upgrade.

What happens when brokers restart

Upgrades that trigger broker restarts have the following impact:

  • Temporary disconnections. Clients may briefly disconnect and reconnect, causing occasional errors.

  • Message send failures. Messages sent during the upgrade may fail. Enable client-side retries to resend failed messages. Messages successfully sent before the upgrade are preserved.

  • Partition reassignment. Services continue, but some messages may be routed to a different partition.

  • Security group updates. Traffic upgrades may scale out the cluster. If new brokers are added, update the client security group to allow access to their IP addresses.

For serverless instances, auto scaling is suspended during upgrades or downgrades. Schedule these changes during stable traffic periods.

Estimated upgrade duration

Instance specification

Estimated duration

alikafka.hr.30xlarge or lower, alikafka.hw.30xlarge or lower

Approximately 30 minutes

alikafka.hr.60xlarge or higher, alikafka.hw.60xlarge or higher

More than 1 hour

Duration increases linearly with the instance specification. A larger disk capacity increase also extends the upgrade time. Schedule upgrades during off-peak hours.

  • Submitting an upgrade request does not affect your existing cluster until payment is completed and the order takes effect. The actual upgrade begins only after the order becomes effective.

  • For subscription instances, the fee incurred by upgrading the instance configuration (such as expanding disk capacity or upgrading the traffic specification) is a one-time price difference. It is not a recurring monthly charge.

Upgrade an instance

  1. Log on to the ApsaraMQ for Kafka console. On the Overview page, select a region in the Resource Distribution section.

  2. On the Instances page, click the name of the target instance.

  3. On the Instance Details page, click Upgrade in the upper-right corner of the Overview section.

  4. In the Instance Upgrade Risks dialog, read the risk description and click OK.

  5. In the configuration change panel, configure the following parameters.

    Parameter

    Description

    Instance Edition

    Supports upgrading Standard Edition to Professional Edition. Note the following when upgrading from Standard Edition to Professional Edition:

    • The disk capacity remains unchanged after the upgrade. For example, if the original disk capacity is 1,000 GB, it remains 1,000 GB after the upgrade.

    • Backup storage capacity is no longer included as a billing item after the upgrade.

    • After the upgrade, you can configure message retention duration at the individual topic level.

    • This upgrade is irreversible. Downgrading from Professional back to Standard is not supported. If you upgrade the configuration directly and the console returns the error message "can not change specType", the instance does not support an in-place upgrade from Standard to Professional. In this case, use the following migration solution instead:

      1. Create a new Professional instance.

      2. Use Cluster Linking or a self-built migration tool to migrate data from the original instance to the new instance.

      3. After you switch your business traffic to the new instance, release the original Standard instance.

    Supports upgrading the traffic specification of Professional Edition to a higher traffic specification. Does not support downgrading Professional Edition to Standard Edition. Does not support mutual upgrades between Open Source Edition (local disk) and other specification types. For information about instance specification types and pricing, see Billing overview.

    Network Type

    Change between VPC and Internet and VPC.

    Public Traffic

    Internet traffic includes read and write traffic with equal maximum throughput. Select a specification based on your higher peak value (read or write). Billed only for Internet-and-VPC instances.

    Note

    ApsaraMQ for Kafka estimates optimal bandwidth based on the instance type. Purchase the Internet traffic as prompted on the buy page to avoid throttling.

    Traffic Specification

    Traffic specification details.

    Disk Capacity

    Default disk capacity varies by traffic specification and adjusts automatically when the specification changes. Modify as needed. If upgrading the traffic specification triggers a cluster scale-out (for example, from 3 nodes × 30 MB/s to 3 nodes × 60 MB/s), the storage capacity of the new nodes must be aligned with the existing nodes, so the disk capacity is doubled accordingly. If the current disk capacity is lower than the minimum value recommended by the system, you must increase it above the recommended value before you can complete the upgrade.

    Partitions

    Total partitions = purchased partitions + free partitions included in the specification. Available topics = total partitions. Available consumer groups = total partitions × 2.

    When you upgrade the traffic specification (for example, to alikafka.hr.16xlarge), the system automatically grants the number of partitions corresponding to the new specification (for example, 2,000 partitions). You do not need to manually adjust the partition count or purchase additional partitions.

    Note

    Since August 26, 2022, upgrading instances purchased by topic specification is no longer supported. Topic, group, and partition quotas of an instance purchased by topic.

  6. Read and select the terms of service, then click Buy Now.

Traffic specification details

Traffic specification upgrades depend on the instance edition and disk type.

Traffic limits by edition:

Edition

Peak throughput

Standard Edition

Up to 120 MB/s read or write

Professional Edition (High Write)

Up to 2,000 MB/s write

Professional Edition (High Read)

Up to 3,000 MB/s read, up to 1,000 MB/s write

Note

To select a traffic specification above 120 MB/s peak read or write rate, upgrade to Professional Edition first.

Cluster scale-out by disk type:

Some traffic upgrades trigger a cluster scale-out. After scale-out, rebalance topic traffic. Rebalance topic traffic.

Disk type

Scale-out trigger

Ultra disk

Traffic specification with peak read or write rate up to 120 MB/s

SSD

Traffic specification with peak read or write rate up to 300 MB/s

Verify the upgrade

After you place the upgrade order, the effective time appears on the order page.

In the Basic Information section of the Instance Details page, the Status changes to Upgrading. After the upgrade completes, the page displays the new configurations.

FAQ

Can I upgrade only the public network bandwidth without upgrading the instance specification?

Yes. For non-Serverless instances, you can go to the Upgrade panel in the console and adjust only the Internet Traffic parameter to increase the public network bandwidth without changing the instance specification.

Note

This operation is an upgrade and triggers a rolling restart of the cluster. During the restart, clients may experience brief disconnections and reconnections, and partition message delivery may temporarily become out of order. Perform this operation during off-peak business hours.

Why isn't the Endpoints option shown in the console?

If the Endpoints option is not displayed on the instance details page, this is usually because the instance is a VPC-connected instance that has not enabled public network access, or the instance resides in a region that does not support public network access, such as Alibaba Finance Cloud regions. To obtain a public endpoint, Upgrade described in this topic and change the network type to Internet and VPC so that the instance becomes an Internet- and VPC-connected instance.

Why does the bandwidth usage on a node increase after I upgrade the public network bandwidth?

This is expected behavior and can occur for the following reasons:

  • Metric scope: Public network bandwidth usage is measured based on the elastic IP address (EIP), which includes management traffic such as heartbeats and offset commits, while service traffic metrics count only message sends and receives.

  • Unit difference: Public network bandwidth is measured in bit/s, while service traffic is measured in B/s. Because 1 byte equals 8 bits, the two metrics can differ by a factor of 8 even for the same amount of traffic.

  • Increased consumption capacity: After the bandwidth limit is raised, clients can consume messages faster, which increases actual bandwidth usage.

  • Higher burst tolerance: After you purchase more bandwidth, the server-side tolerance for traffic bursts that exceed the purchased specification also increases, which can result in higher observed usage during traffic spikes.