SET

Property

Description

Valid values

Default

odps.job.workers.limit

Maximum number of concurrent workers for an entire job.

1–1000000

—

odps.stage.mapper.mem

Memory allocated to each Map worker.

256–12288

1024 MB

odps.stage.reducer.mem

Memory allocated to each Reduce worker.

256–12288

1024 MB

odps.stage.joiner.mem

Memory allocated to each Join worker.

256–12288

1024 MB

odps.stage.mem

Total memory for all workers in a job. Lower priority than odps.stage.mapper.mem, odps.stage.reducer.mem, and odps.stage.joiner.mem.

256–12288

—

odps.stage.mapper.split.size

Input data size per Map worker (split size of the input file). Adjusting this value indirectly controls the number of Map workers.

—

256 MB

odps.stage.reducer.num

Number of workers at each Reduce stage.

1–10000

—

odps.stage.joiner.num

Number of workers at each Join stage.

1–10000

—

odps.stage.num

Number of concurrent workers per stage in a MaxCompute job. Lower priority than odps.stage.mapper.split.size, odps.stage.reducer.num, and odps.stage.joiner.num.

1–10000

—

Property

Description

Valid values

Default

odps.sql.reshuffle.dynamicpt

Enables dynamic partitioning, which prevents a large number of small files from being generated. Set to False when a small number of dynamic partitions are expected, to prevent data skew.

True / False

—

odps.sql.type.system.odps2

Enables the MaxCompute V2.0 data type edition.

True / False

—

odps.sql.hive.compatible

Enables the Hive-compatible data type edition. Required for Hive syntaxes such as inputRecordReader, outputRecordReader, and Serde.

True / False

—

odps.sql.executionengine.batch.rowcount

Number of rows the SQL engine processes per batch. Reduce this value if an out-of-memory (OOM) error occurs due to large row fields, or if UDF performance causes timeouts. Avoid setting this too low, as it degrades SQL engine performance.

1–1024

1024

odps.sql.groupby.position.alias

Treats integer constants in a GROUP BY clause as column positions in a SELECT statement.

True / False

—

odps.sql.groupby.skewindata

Enables the anti-skew feature for GROUP BY operations.

True / False

—

odps.sql.orderby.position.alias

Treats integer constants in an ORDER BY clause as column positions in a SELECT statement.

True / False

—

odps.sql.groupby.orderby.position.alias

Treats integer constants in both GROUP BY and ORDER BY clauses as column positions in a SELECT statement. Setting this to True while jobs are running may affect job parsing. Verify that existing jobs still work correctly before enabling this for a project.

True / False

—

odps.sql.python.version

Python version used when executing SQL statements.

cp27 / cp37

—

odps.sql.timezone

Time zone of the MaxCompute project. For more information, see Time zone configuration operations.

—

—

odps.sql.decimal.tostring.trimzero

Removes trailing zeros after the decimal point for DECIMAL type values.

True / False

True

odps.sql.select.output.format

Controls whether the MaxCompute client (odpscmd) displays column headers in query results. For more information, see MaxCompute client (odpscmd).

{"needHeader":false,"fieldDelim":""} (no headers) / {"needHeader":true,"fieldDelim":""} (with headers)

—

odps.sql.type.json.enable

Enables JSON data types. For more information, see Instructions for using JSON data types in MaxCompute.

True / False

—

odps.sql.executionengine.enable.string.to.date.full.format

Converts date strings that include hour, minute, and second parts.

True / False

False

odps.sql.executionengine.enable.rand.time.seed

Controls the seed used by the RAND function. True uses the current system time (results may differ on rerun; cannot be used as a shuffle key). False uses the current instance ID (ensures idempotency).

True / False

False

odps.sql.metering.value.max

Upper limit on resources consumed by a SQL statement. For more information, see Consumption control.

—

—

odps.sql.job.max.time.hours

Maximum duration a single job can run.

1–72

24 hours

Property

Description

Valid values

Default

odps.sql.mapjoin.memory.max

Memory size for small tables loaded into memory during a MAPJOIN operation. The global memory limit for all small tables in the same task equals min(8 GB, odps.sql.mapjoin.memory.max × number of small tables).

0–8192 MB

—

odps.sql.skewjoin

Enables the SKEWJOIN feature to mitigate long-tail issues.

True / False

—

odps.sql.skewinfo

Specifies the key and value on which SKEWJOIN operates. For more information, see Data skew tuning.

—

—

Property

Description

Valid values

Default

odps.optimizer.enable.range.partial.repartitioning

Enables the Shuffle Remove feature for range-clustered tables.

True / False

—

odps.optimizer.skew.join.topk.num

Number of hot key values the optimizer retrieves during an Aggregate operation. For more information, see SKEWJOIN HINT.

—

—

odps.optimizer.stat.collect.auto

Enables the Freeride feature, which automatically collects column statistics for tables. For more information, see Optimizer.

True / False

—

odps.optimizer.stat.collect.plan

Collection plan for gathering column statistics on specific columns. For more information, see Optimizer.

—

—

Property

Description

Valid values

Default

odps.sql.udf.jvm.memory

Maximum Java Virtual Machine (JVM) heap memory for a user-defined function (UDF). Increase this value if UDF workloads cause OOM errors, but also optimize UDF code to address the root cause.

256–12288

1024 MB

odps.function.timeout

Timeout period for a UDF.

0–3600

600 seconds

odps.sql.udf.getjsonobj.new

Controls whether the GET_JSON_OBJECT function returns the original string as-is. Projects created after April 20, 2021 return original strings by default. Projects created before that date use escape characters for JSON-formatted reserved characters, to avoid affecting existing jobs. For more information, see GET_JSON_OBJECT.

True / False

—

odps.sql.udf.ppr.deterministic

Enables partition pruning for UDFs. For more information, see the "WHERE clause (where_condition)" section in SELECT syntax.

True / False

—

odps.sql.udf.ppr.to.subquery

Ignores errors during result backfilling when partition pruning is performed. For more information, see the "WHERE clause (where_condition)" section in SELECT syntax.

True / False

—

odps.sql.session.resources

Resources referenced by user-defined types (UDTs). Separate multiple resources with commas. For more information, see Overview.

Uploaded resources

—

odps.sql.udt.display.tostring

Enables the java.util.Objects.toString() mechanism on the Wrap for all columns whose output is a UDT.

True / False

—

odps.sql.session.java.imports

Java packages referenced by UDTs. Separate multiple packages with commas. For more information, see Overview.

Uploaded Java packages

—

Property

Description

Valid values

Default

odps.sql.executionengine.coldata.deep.buffer.size.max

Cache size for storing columns with complex data types (such as STRING, MAP, STRUCT, ARRAY, or BINARY) during table writes. For a table with three complex-type columns, MaxCompute reserves 192 MB (64 × 3) of memory by default. If the output table contains variables of complex types, or if a small table involved in a MAPJOIN operation contains data of complex types, adjusting this property may adversely affect the memory used to run jobs. A large value of this property may cause an OOM error. If a large value causes OOM errors, reduce this property based on actual variable sizes and the batch row count setting.

Unit: bytes

67108864

Property

Description

Valid values

Default

odps.sql.external.net.vpc

Enables virtual private cloud (VPC) for external tables. For more information, see HBase external tables (ApsaraDB for HBase Standard Edition or Performance-enhanced Edition), RDS MySQL external tables, or Hologres external tables.

True / False

—

odps.sql.unstructured.data.oss.use.https

Enables HTTPS when accessing Object Storage Service (OSS) external tables at the underlying layer. For more information, see Create an OSS external table.

True / False

—

odps.sql.unstructured.oss.commit.mode

Enables multipart upload to write data to OSS external tables. For more information, see Write data to OSS.

True / False

—

odps.sql.unstructured.tablestore.put.row

Calls the PutRow operation to write data to a Tablestore external table. For more information, see PutRow.

True / False

—

odps.sql.unstructured.external.max.dop

Maximum concurrency of workers when accessing external tables.

Positive integer

—

odps.ext.oss.orc.native

Uses the native ORC reader to read tables.

True / False

—

Property

Description

Valid values

Default

console.sql.result.instancetunnel

Enables InstanceTunnel. For more information, see Usage notes.

True / False

—

Property

Description

Valid values

Default

LabelSecurity

Enables LabelSecurity. For more information, see Label-based access control.

True / False

—

CheckPermissionUsingACL

Enables ACL-based access control. For more information, see MaxCompute permissions.

True / False

—

CheckPermissionUsingPolicy

Enables policy-based access control. For more information, see Policy-based access control.

True / False

—

ObjectCreatorHasAccessPermission

Grants object creators access to the objects they create.

True / False

—

ObjectCreatorHasGrantPermission

Allows object creators to grant other users access to their objects.

True / False

—

ProjectProtection

Enables project data protection. For more information, see Project data protection.

True / False

—

odps.output.field.formatter

Specifies the dynamic data masking rule applied to SQL query results. For more information, see Dynamic data masking.

Custom masking rule

—

odps.isolation.session.enable

Enables dynamic data masking for a single session. For more information, see Dynamic data masking.

True / False

—

Property

Description

Valid values

Default

odps.task.wlm.quota

Name of the quota used to run tasks in the current session. For more information, see Manage quotas in the new MaxCompute console.

—

—