[GitHub] [druid] himanshug commented on issue #10193: Balancer can work incorrectly in case of slow historicals or large number of segments to move

[GitBox <gi...@apache.org>]

himanshug commented on issue #10193:
URL: https://github.com/apache/druid/issues/10193#issuecomment-790941778


   @clintropolis
   
    It is fun to revisit your own old comments and reverse engineer them :)
   
    IIRC, my thought process was, If historicals overall (not single historical node) are not able to keep up with rate of load requests assigned to them and that is causing the timeouts, then the slowness needs to be addressed by e.g. increase number of threads or add more historicals because even if balancer assigns the same segment to another historical then that other historical might also timeout load requests assigned to it.... and this can continue to go on.
   Hence the suggestion to fix the cause of slowness, unless it was one off historical that was slow and the proposed fix would solve the problem for that case.
   
   also, change in #10213 looks fine from the perspective of accurately tracking historical's capacity utilization and consequently not over-allocating segments to it. I don't think I was against having the change made but was probably proposing to address the slowness if it was a cluster level issue.
   
   
   


----------------------------------------------------------------
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:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org