capistrant commented on pull request #11429:
URL: https://github.com/apache/druid/pull/11429#issuecomment-931669321


   > @capistrant This seems cool! For my own curiosity (and maybe as a 
practical example for the docs), Could you explain why it would be beneficial 
to have the segmentMetadata queries at a higher priority than the other queries 
running on the cluster.
   
   What it comes down to is an operators opinion though. Perhaps, operators  
would not want to give these queries a higher priority, and I think that is 
just fine. The flexibility to choose is really the goal here.
   
   My personal opinion is that these broker generated metadata queries are of 
utmost importance and should never be waiting in line behind other queries 
within a historical node. I think this because I always want my brokers to have 
the freshest metadata possible, and therefore I want the queries to achieve 
that freshness to run with a high priority relative to other queries being 
submitted. For instance, in one of the production clusters I help run, our 
services layer that submits queries for users assigns them priorities of 0, 10, 
20, or 30. In this case, keeping a default of 0 for the broker generated 
metadata queries would put them on par with our known "slow queries" 
(relatively speaking) when it comes to prioritization on historicals. Whereas 
we think they should be 30, in step with the most important queries on the 
cluster.


-- 
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