vaughn

## Question 1. 关于幂等看了 历史 issure, 结合tcc代码，目前框架中应该是没有做幂等处理，然后定时恢复任务有用 lock 函数，但看了 redis 的实现感觉无法做到幂等，所以想跟大佬们确认下幂等的问题应该是要业务上自己处理对吧? 2. 假如 try 函数执行成功了，但是更新 trying 状态的时候down了，这个时候貌似走不到恢复（pre_try 会跳过）,定时任务会清理 trans, 不会走 cancel，但是这个场景应该走 cancel, 请问大佬们这个极端场景问题是否存在？

### Feature Description (功能描述) Now, hugegraph cluster node roles (now used by task-scheduler) are created by config file when the startup, this is complex for the OP, so here we...

### Feature Description (功能描述) Now, Hugegraph supports a multi-graph feature, but some resource is stored in thread-local, so that resource don't be released when the graph be removed or closed.