MaxCompute
Connect Alibaba Cloud MaxCompute (ODPS) for analysis. MaxCompute is queried natively: the SQL is pushed down to the MaxCompute engine and only the result is returned. This suits very large tables that are impractical to mount into a local engine.
When to use
- Large-scale analytical tables held in MaxCompute.
- Reporting over partitioned MaxCompute tables where the computation should stay in MaxCompute.
Because it is a native connector, MaxCompute datasets are queried in MaxCompute's own SQL dialect and cannot be joined with datasets from other connectors in a single query.
Before you begin
Prepare an Alibaba Cloud AccessKey (ID and secret) belonging to a RAM user with:
- Read access to the target MaxCompute project and its tables.
- Membership in the project (a RAM user that is not a project member cannot list or read its tables).
Grant the least privilege that allows reading the tables you will analyze.
Connection settings
| Field | Required | Description |
|---|---|---|
| Endpoint | Yes | The MaxCompute service endpoint for your region, e.g. http://service.cn-hangzhou.maxcompute.aliyun.com/api. |
| Project | Yes | The MaxCompute project name. |
| AccessKey ID | Yes | The AccessKey ID of the RAM user. |
| AccessKey Secret | Yes | The AccessKey secret. Stored encrypted; never shown after saving. |
Test and save
Enter the fields and use Test connection. The test verifies that the project exists and is accessible with the provided credentials, not merely that the endpoint responds. On success, choose the tables to import.
Capabilities
- Execution model: Native. Queries run in MaxCompute; results are returned to Datyo. Cross-source joins are not available.
- SQL dialect: MaxCompute (ODPS) SQL. The Agent generates SQL in this dialect for MaxCompute datasets.
- Access: Read-only analysis.
Working with partitioned tables
MaxCompute commonly enforces partition pruning on large tables. When analyzing a partitioned table, constrain the query to specific partitions (for example a date range) rather than scanning all partitions, both for cost and to satisfy the source's guards. Describe the partitioning of a table in its column descriptions or in the Space's business notes so the Agent includes appropriate partition filters.
Troubleshooting
- Connection reports success but queries fail — confirm the RAM user is a member of the project and can read the specific table, not just that the AccessKey is valid.
- Permission denied — the AccessKey lacks read access to the table; grant it at the project or table level.
- Full-scan rejected — add a partition filter to the analysis; unbounded scans of partitioned tables are blocked by the source.
- Wrong region — set Endpoint to the region where the project resides.