Spark3.2.0通过JDBC连接SpliceMachine查询长度受限报错

数据老陈

2026-8-4

问题重现

我们使用Spark 3.2.0通过JDBC连接Splice Machine数据库拉取数据，驱动版本为com.splicemachine:db-client:2.7.0.1815，代码示例如下：

val url_upd = "jdbc:splice://sl73caehdp0225.visa.com:1527/splicedb;user=***;password=a***"
val query = "select top 5 ENCRPT_PYMT_CRD_ACCT_NUM_NORM from optd.ttd_cs_dtl CS where TRAN_CD IN ('01','04','05','06','10','11')"

val jdbcDF = spark.read.format("jdbc")
  .option("url", url_upd)
  .option("driver", "com.splicemachine.db.jdbc.ClientDriver")
  .option("query", query)
  .load();

当查询语句长度超过111字符时，会抛出如下错误：

"A connection could not be established because the database name '' is larger than the maximum length allowed by the network protocol."

相同的连接配置和查询语句在Spark 2.3中可正常运行。

解决方案

1. 用`dbtable`参数替代`query`，将查询包装为子查询

Spark 3.x对JDBC的query参数处理逻辑和2.3存在差异，会导致Splice驱动误解析参数内容。将查询语句包装成子查询，通过dbtable参数传递可绕过该问题：

val url_upd = "jdbc:splice://sl73caehdp0225.visa.com:1527/splicedb;user=***;password=a***"
val subQuery = "(select top 5 ENCRPT_PYMT_CRD_ACCT_NUM_NORM from optd.ttd_cs_dtl CS where TRAN_CD IN ('01','04','05','06','10','11')) as temp_table"

val jdbcDF = spark.read.format("jdbc")
  .option("url", url_upd)
  .option("driver", "com.splicemachine.db.jdbc.ClientDriver")
  .option("dbtable", subQuery)
  .load();