dm 任务跑某张表，但是turncate 的语句总是被跳过，找不到配置的方法

有猫万事足 · 2023 年9 月 26 日 04:08

pingcap/tiflow/blob/v5.3.0/dm/syncer/syncer.go#L2234


      
          	}
          	qec.tctx.L().Info("resolve sql", zap.String("event", "query"), zap.Strings("appliedDDLs", qec.appliedDDLs), zap.Stringer("queryEventContext", qec))
          
          	metrics.BinlogEventCost.WithLabelValues(metrics.BinlogEventCostStageGenQuery, s.cfg.Name, s.cfg.WorkerName, s.cfg.SourceID).Observe(time.Since(qec.startTime).Seconds())
          
          	/*
          		we construct a application transaction for ddl. we save checkpoint after we execute all ddls
          		Here's a brief discussion for implement:
          		* non sharding table: make no difference
          		* sharding table - we limit one ddl event only contains operation for same table
          		  * drop database / drop table / truncate table: we ignore these operations
          		  * create database / create table / create index / drop index / alter table:
          			operation is only for same table,  make no difference
          		  * rename table
          			* online ddl: we would ignore rename ghost table,  make no difference
          			* other rename: we don't allow user to execute more than one rename operation in one ddl event, then it would make no difference
          	*/
          
          	qec.needHandleDDLs = make([]string, 0, len(qec.appliedDDLs))
          	qec.trackInfos = make([]*ddlInfo, 0, len(qec.appliedDDLs))

从报错的位置看，你的dm版本是5.3.0。然后上面写了一大段注释。

		* sharding table - we limit one ddl event only contains operation for same table
		* drop database / drop table / truncate table: we ignore these operations

可能是有sharding表的聚合存在，所以上游单个truncate table并不执行，而是需要全部shard group的表都做这个ddl，下游才会truncate。

不过单看你这个任务怎么看也不是一个sharding table。是有其他的任务也在同步不同的上游库进下游的同一张表？

https://github.com/pingcap/tiflow/issues/5055
github上的类似issue。

论坛相关的讨论