Help Center/ Data Lake Insight/ FAQs/ Problems Related to SQL Jobs/ Job O&M Errors/ Why Is Error "IllegalArgumentException: Buffer size too small. size" Reported When Data Is Loaded to an OBS Foreign Table?
Updated on 2023-05-19 GMT+08:00

Why Is Error "IllegalArgumentException: Buffer size too small. size" Reported When Data Is Loaded to an OBS Foreign Table?

Symptom

The following error message is displayed when the LOAD DATA command is executed by a Spark SQL job to import data to a DLI table:

error.DLI.0001: IllegalArgumentException: Buffer size too small. size = 262144 needed = 2272881
In some cases ,the following error message is displayed:
error.DLI.0999: InvalidProtocolBufferException: EOF in compressed stream footer position: 3 length: 479 range: 0 offset: 3 limit: 479 range 0 = 0 to 479 while trying to read 143805 bytes

Possible Causes

The data volume of the file to be imported is large and the value of spark.sql.shuffle.partitions is too large. As a result, the cache size is insufficient.

Solution

Decrease the spark.sql.shuffle.partitions value. To set this parameter, perform the following steps:

  1. Log in to the DLI management console and choose Job Management > SQL Jobs. In the Operation column of the target SQL job, click Edit to go to the SQL Editor page.
  2. On the displayed page, click Set Property and set the parameter.
    Figure 1 Modifying the parameter
  3. Execute the job again.