[
https://issues.apache.org/jira/browse/FLINK-28345?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17561418#comment-17561418
]
Feng Jin commented on FLINK-28345:
----------------------------------
Whether it will increase the complexity of generic JDBC implementation is worth
considering, but I don't think adding this judgment is a special
implementation, or maybe a general optimization ? Maybe I have some aspects I
didn't take into account, please correct me if I'm wrong. In view of the fact
that the community officials have not added support for the clickhouse system,
most users may not encounter this problem, but at least I hope this scenario
can be considered in the new version of the connector.
> Flink JDBC connector should check batch count before flush
> ----------------------------------------------------------
>
> Key: FLINK-28345
> URL: https://issues.apache.org/jira/browse/FLINK-28345
> Project: Flink
> Issue Type: Improvement
> Components: Connectors / JDBC
> Affects Versions: 1.15.0, 1.14.5
> Reporter: Feng Jin
> Priority: Major
>
> org.apache.flink.connector.jdbc.internal.JdbcOutputFormat#flush
> {code:java}
> //代码占位符
> @Override
> public synchronized void flush() throws IOException {
> checkFlushException();
> for (int i = 0; i <= executionOptions.getMaxRetries(); i++) {
> try {
> attemptFlush();
> batchCount = 0;
> break;
> ....{code}
> When flush the batch, we should check batchCount is grater than 0. Other
> wise it would cause some problem with some drivers that do not support empty
> batches, like clickhouse jdbc driver.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
