symious commented on PR #5603:
URL: https://github.com/apache/hadoop/pull/5603#issuecomment-1533942745

   > Well, I'm afraid your specific problem does not match Dee why do use cases 
of uploading to stores without checksums. Now, I would I've been happier if 
distcp's -skipCrc option was required to copy data from an FS with checksums to 
one without, but it is not and to add it now would break so many people's 
workflows.
   > 
   > So what do we do here?
   > 
   > maybe: create counters of why files were copied, specifically
   > 
   > * not found at destination
   > * file length different
   > * modtime
   > * checksum
   > 
   > Then after a job you can see why files were copied from the host where the 
job was launched. Then if you want to know why there were issues such as 
checksums and modtimes, you can log out to debug. Obviously, this will be 
something to add to the distcp documentation.
   > 
   > Now: big warning. I am personally scared of distCp. It is a critical 
workflow tool and even use programmatically, yet it is surprisingly brittle. It 
is a running joke that's the last person two add any code to the module gets to 
field or support calls until someone else comes along. Thank you for 
volunteering! This also explains why we will be very reluctant/strict about 
taking on changes. Don't take it personally is as hey everyone gets that same 
grilling here.
   
   Totally understand. 
   
   I have removed the error log if checksum is null for sourcePath and 
targetPath, so a FS without checksum won't be affected.
   
   I have also changed the error log to INFO level if both checksum exists but 
not equal, or a DEBUG level will be more preferred? Personally I think INFO 
level can save us more time since we don't need to reconfig and restart the 
job. 


-- 
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: common-issues-unsubscr...@hadoop.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org

Reply via email to