jihoonson commented on pull request #10920:
URL: https://github.com/apache/druid/pull/10920#issuecomment-883624912


   > > @jihoonson could you create a branch in the druid repo for me please?
   > 
   > @JulianJaffePinterest Here you go - 
https://github.com/apache/druid/tree/spark_druid_connector
   
   Sorry @JulianJaffePinterest, I missed your comment. @abhishekagarwal87 thank 
you for creating a branch.
   
   > @jihoonson I don't have the ability to create a branch in this repo to 
open prs against, but I'm game to try. It's been difficult to get reviews from 
committers (it took two months and multiple emails to the dev list to get the 
first review for this diff) and so I'm hesitant to go down a path of needing to 
hunt reviews for many diffs, but if you think smaller prs will actually get 
reviews I can do so. (As an aside, if the Druid website's [list of 
committers](https://druid.apache.org/community/) is accurate only 1 of the last 
10 prs tagged "Design Review" actually had approval from three committers in 
GitHub, so some process improvements may be necessary)
   
   I'm fine with either working on a branch or merging this pr directly. 
Whatever the fastest way for merging this change would be best to me. Since 
@samarthjain already reviewed this PR, you need 2 more approvals from 
committers. So, it's up to you which way you choose. I suggested working on a 
branch because committers are usually busy and big PRs usually hard to get 
reviews.
   
   For the design review process, I believe we are doing better than you 
described. I checked some of last PRs tagged design review, and they do have 3 
approvals from committers, including the PR author himself. Perhaps the 
committer list on our website has gone stale. You can find the most recent list 
of committers in https://people.apache.org/phonebook.html?unix=druid. (We 
probably should add this link in our website instead of our own committer list)
   
   > For documenting level of confidence in the various deep storage/metadata 
options, would you prefer an explicit table in the docs or in-line callouts 
(the current approach). 
   
   Oh, I didn't realize they are called out! Thank you for adding them. Perhaps 
I didn't read the doc carefully. But it might mean that it could be better if 
the callout was more obvious or more emphasized. I'm fine with either a table 
or callouts, but prefer whatever the most explicit way.
   
   > I haven't updated .travis.yml beyond the checkstyle because the existing 
travis packages run the static checks and tests for the new module. In addition 
to unit tests, there's a light end-to-end test as well that writes segments out 
and then reads the created segments and confirms they match what was written. 
If there's actually a road for this pr, I can look into writing an integration 
test that requires an external spark cluster but obviously that won't be run as 
part of travis.
   
   These tests sound really nice. I think it's OK to not have integration tests 
or not run those unit tests automatically in this PR, but we should do it as 
soon as possible. Do you have a follow-up plan for this?


-- 
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.

To unsubscribe, e-mail: [email protected]

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



---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to