Details
-
Improvement
-
Status: Open
-
P3
-
Resolution: Unresolved
-
None
-
None
-
None
Description
Currently we have a number of great release related scripts that are mentioned in Beam release guide.
https://beam.apache.org/contribute/release-guide/
These tools are great but I think given the length of these scripts, we should make them more resilient to error conditions and make them retry commands (after user action) as much as possible.
For example, RC creation script might fail for a number of reasons (it failed for me due to tox not being available for example) and halt at an intermediate state (some files staged, some commits done to Github but not complete). Re-running script fails due to previous state. After this it is up to the release manager to figure out where the script failed, do cleanups, and resume the process.
Also, have we considered using some release automation tooling instead of shell scripts ?
boyuanz assigning this to you since you developed some of these. Please feel free to re-direct.
cc: altay