hi It has been fixed, see more details in this JIRA: https://issues.apache.org/jira/browse/KYLIN-3562.
you Zhuang <zhuangzixiao...@gmail.com> 于2018年12月5日周三 上午4:02写道: > org.apache.kylin.common.persistence.WriteConflictException: > Overwriting conflict /user/admin_mock, expect old TS 1543976358968, but it > is 1543976359060 > > I dive into *org.apache.kylin.rest.service.KylinUserService* class, the > *updateUser* method *setEvictCacheFlag(true)*; > > After *getKylinUserManager().update(managedUser)*; > > I suspect that next login will not hit cache, and the inner real > authentication provider always will authenticate. > > > And the other place I think that is not reasonable > is getKylinUserManager().update(managedUser); > > Method, the *user.setLastModified(exist.getLastModified());* line is not > reliable in cluster mode because of constant update described above. > > > Which depends network notification to update inner user map cache. When > I comment setEvictCacheFlag(true); > > And add* exist = crud.reloadQuietly(user.getUsername());* line , seems > everything is going right. > > > > -- Regards! Aron Tao