提供的日志片段里可以看到用的是incr的同步模式，是有drop的DDL执行，而且batcher也设置了barrier:

[2025/12/08 14:45:09 CST] [INFO] Syncer[rs_xiaobinsource] checkCheckpointUpdate find barrier
...
[2025/12/08 14:45:09 CST] [INFO] Replay-0 run DDL with metadata[ycsb] in db[{ <nil>}], firstLog: {"ts":{"T":1765176178,"I":1},"t":2,"h":0,"v":2,"op":"c","ns":"ycsb.$cmd","o":[{"Key":"drop","Value":"xiaobintest"}],"o2":[{"Key":"numRecords","Value":3035219}],"PrevOpTime":null,"ui":{"Subtype":4,"Data":"6P+WX4YbR7S/vNZKYRBmLA=="}}
[2025/12/08 14:45:09 CST] [INFO] RunCommand run DDL with type[drop]
[2025/12/08 14:45:09 CST] [INFO] run applyOps for drop: map[ns:ycsb.$cmd o:[{drop xiaobintest}] o2:[{numRecords 3035219}] op:c ts:{1765176178 1} ui:0xc036615900 v:2]

从后续的日志来看，拉取的源端在drop后还有其他DML的oplog:

[2025/12/08 14:45:09 CST] [INFO] Replay-5 Executor-5 doSync oplogRecords received[1024] merged[2]. merge to 0.20% chunks

问一下，ycsb load/run脚本是持续运行中，然后在另外一个会话里执行了dropCollection的操作么？最后的结果具体差异表现是什么样的？
我也会尝试做一些复现；当然，如果你能提供关于复现步骤的更多信息最好。

Uh oh!

DDL 和DML操作没有按照oplog顺序同步，导致源库和目标库数据不一致 #934

Description

Activity

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions