Meng (Ethan) Li

> I think a good reason to catching this early (i.e. at submission time) is that the cycle time for error detection is longer. It is not hard to find...

Since it is possible to break backwards compatibility, I am more conservative on this and would like more tests to be done. (1) Did we get a chance to test...

(1) For old clients, I am thinking of users using old storm client to submit topologies to upgraded cluster. (2) For old python code, I am thinking of * topologies...

> * I did comprehensive search: grep get_msg/set_msg on **/_.py, **/_.java, **/*.clj. All calls are accounted for this change request. > * So this leaves only the user topology jars,...

> @bipinprasad As discussed, I feel all the thrift API methods thould wrap these exception under `TException`. So we can consistently ensure message/cause is passed down to the client side....

Can we modify TopologySummary instead of adding a new structure?

@wxy929629 Thanks for contributing! Can you please help me understand this problem better? Do you mean when Kafka broker is set up with `SASL_PLAIN`, missing `sasl.mechanism` will lead to failures...

Thanks for your contribution. Please file a JIRA and then update this PR **title** and **commit message** to prefix with the JIRA number.

Please also update your commit message to prefix with the JIRA number. Thanks. Other than that, I am +1

Sorry I have a second thought on this. Although `serialVersionUID` is recommended, we should not add it blindly. `serialVersionUID` should be added where there is a need for backwards compatibility...