pipeline-service
pipeline-service copied to clipboard
Improve UX of installation/setup scripts
Currently, there are many tradeoffs and missing scenarios in the installation and setup scripts. This issue will track the overall progress of improving the user experience. Please update the list as needed.
- [ ] Clean the containers in case of failure
- [ ] Stop running kcp processes in case of failure
- [ ] Add parameter/environment variable to enable compilation of kcp binaries when not found
Note: This is for local development environment.
Some other notes from my "local development" experiences:
- kcp in a pod requests a full CPU. This often means that kcp can't be scheduled given that we install a lot of other components that are resource-heavy.
- Our script simply stops when kcp fails to deploy, with no error message.
More notes:
- After overcoming
ckcp
install issues, I struggle with the "what's next" step. Our current README docs aren't exactly clear on this. - The setup scripts default to laying out credentials for KCP and the compute cluster in a temporary directory. Don't developers need these to do any real work? Should we change the location of this data to be somewhere that is more local to the script?
@avinal A lot changed since you opened this issue, and many improvement were made. I'll mark this issue as stale. Please put a comment if you still have pain points that need to be looked after.
Sure @Roming22 I don't think this issue is relevant now. We should close it.
/close
Per @avinal's comment