thenatog edited a comment on pull request #4670:
URL: https://github.com/apache/nifi/pull/4670#issuecomment-730635492


   I've tested out using the diagnosis tool and had some feedback on its usage. 
I've compiled my comments alongside the logs I saw for different tests I tried. 
Let me know if you can't access: 
https://docs.google.com/spreadsheets/d/1kAbM4LLA3NgRjKAfCXg7GqS8dvJnFA6sajrAkBezMt4/edit?usp=sharing
   
   I think with this tool, the key things to focus on should be these:
   
   - The tool is being created to assist users with checking configuration that 
can be difficult to get right. The tool needs to be easy to use. The usage 
guide of the tool wasn't completely clear when I tried out using it.
   - The tool diagnoses errors with configuration, so the tool itself needs to 
be error free (as best as possible). Otherwise if there are errors with the 
tool, how will a user (a user who may already be struggling to get their 
configuration right) know whether the tool is to blame or the configuration is 
to blame? It needs to be pretty robust.
   - Sometimes I got a summary of results, sometimes I didn't, depending on 
what error the tool experienced. It wasn't immediately clear if the tool 
failed, or my configuration did. We need to gracefully fail. In as many cases 
as possible, there should always some form of diagnosis summary at the end of 
running the tool with information on how to proceed/fix the problem. 
   
   I can appreciate that the tls-toolkit code as it stands needs 
redesigning/refactoring, and maybe you can let me know if the above feature 
requests are difficult to implement with the way it is right now. However, we 
should do our absolute best to make sure this diagnosis tool is robust as 
possible. A tool that diagnoses errors should have few errors of its own, and 
its correct usage should be clear and simple. Ideally, we don't want to have to 
provide support to users on how to use a tool that was created to support them 
in the first place.
   
   Having said that, this is a great idea and I think with a few adjustments it 
will be really useful to diagnose problems our users frequently have. 
Especially so if we can do a more complex option for clustered nodes.


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to