I think SQL parse and route to shadow datasource is a better option for using threadlocal. Because of threadlocal only can be used in to sharding-jdbc, the SQL parser way can use into sharding-jdbc and sharding-proxy together, is it?
------------------ Liang Zhang (John) Apache ShardingSphere & Dubbo Zhiyi Yan <[email protected]> 于2019年12月3日周二 下午9:52写道： > Thanks for you suggestion, it is a good idea! > > [email protected] <[email protected]> 于2019年12月3日周二 下午4:56写道： > > > Yes, it is a good design. > > I have a suggestion, mailing list can not display picture, how about > paste > > the picture in to github issue? > > > > ------------------ > > > > Liang Zhang (John) > > Apache ShardingSphere & Dubbo > > > > > > Zhiyi Yan <[email protected]> 于2019年12月3日周二 上午11:08写道： > > > > > Hi ShardingSphere community, > > > > > > We are going to add a new feature - shadow database, this feature is > > > designed to allow dirty data to be written to the shadow database when > > > applying the online environment for pressure measurement, the relevant > > > issue is : > > > https://github.com/apache/incubator-shardingsphere/issues/3269 > > > > > > We want to do this through SQL parse and add router, add a tag field in > > > sql, and according to this field, route to shadow database. > > > > > > Here is a simple design，and we chose the second way. because > ThreadLocal > > > might introduce other problems. > > > > > > [image: sharding-shadow.png] > > > > > > Comments or Suggestions on this design are welcome! > > > > > > Thanks! > > > -- > > > Zhiyi Yan (Zhyee) > > > Apache ShardingSphere > > > > > > > > -- > Zhiyi Yan (Zhyee) > Apache ShardingSphere >
