Tablestore:Why does a HTTP 5XX status code occur when I use Tablestore?

Problem description

Why I use Tablestore, a 5XX status code occasionally occurs. The following table describes the 5XX HTTP status codes.

Cause

Tablestore is a table-based distributed serverless storage service. The Tablestore server automatically performs load balancing based on the data volume and access status of each partition. This way, data storage and access capabilities can be seamlessly scaled regardless of the capability limit of a single server.

Tablestore distributes data to different partitions base on the order of the first primary key column. Different partitions are distributed to different service nodes to provide data reading and writing services.

The dynamic load balancing mechanism of Tablestore can detect whether a partition such as P1 stores large amounts of data or processes too frequent access. In this case, the partition is split into two partitions P1 and P5. The two partitions are distributed to the service node that has lower loads.

Tablestore uses the preceding automatic load balancing mechanism to automatically scale the storage and concurrent access capabilities without human intervention during the whole process.

Tablestore adopts the shared storage mechanism that uses partitions as logical units. Therefore, data is not migrated, but metadata of the data table is changed during load balancing. During the metadata change process, specific partitions may become unavailable to ensure data consistency. In normal cases, the unavailability lasts hundreds of milliseconds. When the workloads of the partitions are heavy, the unavailability lasts seconds. If you perform read or write operations on the partitions during the period of time, the preceding error may occur.

Solution

Retry the operation. Tablestore SDKs provide retry policies. You can specify retry policies when you initialize a Tablestore client.

Tablestore uses the API that complies with the standard Restful protocol. Due to the uncontrollable network environment, we recommend that you add a retry policy for all read and write operations to respond to network errors for fault tolerance.