Master Inceptor: Essential Q&A for Getting Started with This Big Data Engine

Inceptor Q&A

When first encountering Inceptor, many users have basic questions or encounter issues; this article compiles common Q&A to help resolve them.

What is Inceptor? Inceptor is built on the Hadoop platform as an efficient batch‑processing analytical database that solves large‑scale data processing challenges. The community edition fully supports the SQL‑2003 standard and provides standard JDBC/ODBC connections.

Where can I get the Inceptor manual? Visit the Transpedia online documentation service at https://docs.transwarp.cn/ .

How do I interact with Inceptor? Use Beeline as the command‑line tool. Connect with the following command:

beeline -u "jdbc:hive2://<inceptor_server>:10000/<database_name>"

How to create a table and query it?

CREATE TABLE quickstart (a INT, b STRING);
LOAD DATA INPATH '/tmp/quickstart.txt' OVERWRITE INTO TABLE quickstart;
SELECT * FROM quickstart LIMIT 10;

How to handle SQL execution errors? Use the returned error message and code to consult the "Inceptor Error Codes and Information" manual on Transpedia ( https://docs.transwarp.io/ ).

What table types does the community edition support?

External tables (no ownership) and managed tables (ownership).

Storage formats: TEXT, ORC, CSV.

Partitioned vs. non‑partitioned tables.

Bucketed vs. non‑bucketed tables.

What are external and managed tables? External tables store only metadata; data remains unchanged and is not deleted when the table is dropped, making them suitable for query‑only scenarios. Managed tables own both metadata and data; dropping the table removes the data as well.

Partitioning considerations

Prefer range partitioning over single‑value partitioning.

Use date or region fields as partition keys.

Avoid secondary partitions.

Keep the number of partitions between 0 and 200.

Number of buckets per partition should be slightly less than the CPU core count.

Design partitions so a single CPU can process all data in one round.

Limit buckets per partition to under 500 and the product of partitions and buckets to under 10,000.

Bucketing considerations

Choose a field with uniform distribution and low repeat rate, such as a primary key.

Check the size of the largest bucket file to avoid data skew.

Each bucket should be less than 1000 MB before compression, typically containing fewer than 10 million rows.

Since bucket count cannot increase with data growth, incremental tables should use both partitioning and bucketing.

Beeline does not display full column data

Adjust the display width with the following parameters: maxWidth: maximum total width (characters). maxColumnWidth: maximum width per column (characters).

# Set max display width to 1500 characters per column and overall
beeline -u jdbc:hive2://localhost:10000/default --maxWidth=1500 --maxColumnWidth=1500

Inceptor execution modes

Supports local and cluster modes. cluster is suited for batch jobs, while local fits low‑latency, high‑concurrency scenarios with small data volumes. Switch using the ngmr.exec.mode parameter (default cluster).

Address already in use: SparkUI

Port 4040 is occupied; free it because TDH assigns this port to the Inceptor UI.

Could not create server socket on address 0.0.0.0:10000

Port 10000 is occupied; free it because TDH needs this port for the Inceptor Server.

Some executors not started

This usually occurs when available CPU or memory is insufficient for the executor requirements. Check YARN resource settings ( yarn.nodemanager.resource.cpu‑vcores and yarn.nodemanager.resource.memory‑mb) and reduce the Inceptor executor resource allocation accordingly.

How to add, delete, or modify table columns?

ALTER TABLE [table] ADD COLUMNS (col1 TYPE, col2 TYPE, ...);
ALTER TABLE [table] DELETE COLUMNS (col1, col2, ...);
ALTER TABLE [table] CHANGE COLUMN old_name new_name TYPE;

How to view the execution plan of a SQL statement? EXPLAIN SELECT * FROM table1; How to migrate tables from an old cluster to a new one?

On the old cluster, run show create table <dest_tbl>; and record the DDL.

Locate the HDFS directory of the source table (e.g., sudo -u hdfs hdfs dfs -ls /<dir>).

On the new cluster, create the target database.

Identify the active NameNode of each cluster and copy data with

hadoop distcp hdfs://<ActiveNN1>:8020/path/to/<dest_tbl> hdfs://<ActiveNN2>:8020/path/to/<dest_db>

.

On the new cluster, execute the recorded DDL to create the table.

If the table is partitioned, add partitions manually, e.g.,

ALTER TABLE tbl ADD PARTITION (p_key='p_value') LOCATION '<hdfs_partition_address>';

How to remove quotation marks when importing CSV files?

DROP TABLE IF EXISTS csv_table;
CREATE EXTERNAL TABLE csv_table(col1 STRING, col2 STRING, col3 STRING)
STORED AS csvfile
LOCATION '/temp/csv/data'
TBLPROPERTIES ('field.delim'=',', 'quote.delim'='"', 'line.delim'='
');

The quote.delim property specifies the character used to quote fields; setting it removes surrounding double quotes during import.

If you have other questions, visit the StarRing forum at http://support.transwarp.cn or email [email protected] .