Closed henrysun007 closed 1 year ago
The original design is to separate executors with other components to make it more flexible. In this implementation, if I understand correctly, executors are talking to the storage directly. Why not return the logs with the function return value instead?
Keeping the logs persistent is good for audit. The log is more a supplement to TaskState
that is persistent in the storage
service than just a diagnosis
tool for developers here. Besides, the log content may be large so it is not very suitable as part of the result. However, as you have already been used to the result way in the cleanroom
branch, we can use it in the master
. Large log scenarios can be handled in the future.
Description
Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context.
Fixes # (issue)
https://github.com/apache/incubator-teaclave/issues/669
The task logger part. Platform logging will be in another PR in a while.
Type of change (select or add applied and delete the others)
How has this been tested?
Checklist
master
.