Discussion: How should FE display the execution error?

[BowenXiao1999]

Now FE scheduler do not correctly display the execution error, see

https://github.com/singularity-data/risingwave/blob/a50393a30b7830d392b0092af460871c8e9cede1/src/frontend/src/scheduler/distributed/stage.rs#L393

Should frontend aware of the detail reason like (ExprError, Divde by Zero error etc)? If yes, maybe we should consider how to encode this into proto.

cc @liurenjie1024. May relates to #3908

[liurenjie1024]

Currently we don't do fault tolerant of query execution, so I think frontend doesn't need to know the exact error kind, and just the error message.

[fuyufjh]

+1 for error message only, because it seems difficult to define all kinds of errors.

[github-actions]

This issue has been open for 60 days with no activity. Could you please update the status? Feel free to continue discussion or close as not planned.

[fuyufjh]

cc. @TennyZhuang

[BowenXiao1999]

Now the solution is we wrap the execution err msg in string into a Err struct returned by streaming GRPC API. So when there is an execution failure in CN tasks, FE scheduler will imediately know and return Err.

risingwave/src/frontend/src/scheduler/distributed/stage.rs

Line 374 in 6f1764f

let status = stauts_res_inner.map_err(SchedulerError::from)?;

risingwave/proto/task_service.proto

Line 75 in 6f1764f

rpc CreateTask(CreateTaskRequest) returns (stream TaskInfoResponse);

Now FE should be able to get the execution failure so I think we can close this issue now.