向db2数据库入库慢

向db2（10.5 fp8）插入数据，无论数据源是csv还是另一个db2库，rdbmswriter速度都很慢，单个channel约1000rec/s. 数据源为db2时, reader端可以跑满千兆带宽，BatchSize=4096，JVM虚机4G, job配置如下：
```
"writer": {
"name": "rdbmswriter",
"parameter": {
"connection": [
{
"jdbcUrl": "jdbc:db2://********",
"table": ["********"]
}
],
"username": "********",
"password": "********",
"table": "********",
"column": [
"STATIS_MONTH",
"SUBS_ID",
"GRP_SUBS_ID",
"PHONE_NO",
"ISPRIMA",
"START_TIME"
],
"#preSql": [
"delete from ********;"
]
}
}
```
多channel下，基本上等于1000*n rec/s，开8个channel时，偶发ERROR=-4229，然后批量插入失败, 变为逐条插入. 单个channel时日志中未发现此情况.
插入速度实在是太慢了，求教解决方法.

Start at the rdbmswriter implementation and its batch-insert error handling, using the supplied DB2 JDBC configuration, BatchSize=4096, and single-channel versus eight-channel results as the reproduction. Check DB2 ERROR=-4229 and the fallback to row-by-row inserts; done means explaining the throughput limit and correctly handling the reported batch failure, with a regression check if the project provides one.

Written by the indexing model from the issue text.