You can use the specification calculator in the console to calculate the required production specification.
CKafka instances are divided into Standard Edition and Pro Edition according to their specifications. For the comparison between the two editions, see Product Specifications.
You can select an appropriate product specification as needed. For the billing rules of different specifications, see Billing Overview.
The migration modes of the Standard and Pro Editions are lossless, non-stop, and imperceptible to the client.
Standard Edition uses a shared cluster. If the available resources in the cluster are sufficient, instances will be upgraded by increasing the quota, with no data migration involved; therefore, upgrade can be completed in several minutes normally. If the available resources in the cluster are insufficient, the system will prompt that the resources are insufficient, and you need to submit a ticket for data migration and resource allocation on the backend and then perform normal Standard Edition instance upgrade in the console.
Pro Edition uses a dedicated cluster. When the system algorithm finds that the current dedicated resource pool cannot sustain the resource quota after instance upgrade, it will automatically add resources to the resource pool and migrate instance data. Therefore, Pro Edition has two instance upgrade modes:
If data migration is required, you can customize the migration time:
If data migration is required, you can select a migration mode and set scheduled migration. There are two migration modes:
As migration will use cluster bandwidth and disk resources, you may worry about that the migration will affect the business during peak hours. In this case, you can set scheduled migration; for example, you can set the start time of instance upgrade to midnight. For detailed directions of instance upgrade, see Upgrading Instance.
Currently, CKafka supports open-source Snappy and lz4 message compression formats. Gzip compression is not supported currently as it requires high CPU usage. If compression is enabled, it may increase the server pressure in some cases; for example, if the producer protocol version is high, but the consumer protocol version is low, downward version conversion may occur during consumption, increasing the server pressure. Therefore, we recommend you disable message compression before performing a stress test.
Configuration Method: In the configuration file of the producer, set the
compression.type parameter to
lz4. The default value is
none, indicating that the feature is disabled.
We recommend you preferentially select Snappy for compression. After compression is enabled, the server CPU utilization and the production and consumption time may increase. Therefore, you need to enable it with caution.