seata生成反向SQL时序列化较慢，有没有什么办法能够提升效率

[Rayxxx]

服务部署：K8S部署，宿主机CPU为8核，部署了15个pod
问题场景：ESP场景，数据量大概为2M，执行update操作时，生成反向SQL大约耗时1.5S左右

[slievrly]

How many rows of records?

[Rayxxx]

30 to 50

[funky-eyes]

Are you certain the UPDATE statement is using an index during execution?

[Rayxxx]

yes,I update using the unique index.

[funky-eyes]

Could you use Arthas to trace the actual execution time under non-concurrent conditions?

[Bughue]

Hi @Rayxxx — any progress on your side since the last replies (e.g., the Arthas trace under non-concurrent conditions)?
We’re hoping you can still reproduce the ~1.5s reverse-SQL case so we can discuss the details together.
If possible, please share Seata client/server + JDK + DB/driver versions and a minimal repro (DDL + typical UPDATE).
Thanks — we’ll follow up once we can align on the same setup.

[Bughue]

#7984