[GitHub] [hadoop] tasanuma commented on a change in pull request #2240: HADOOP-17165. Implement service-user feature in DecayRPCScheduler.

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

tasanuma commented on a change in pull request #2240:
URL: https://github.com/apache/hadoop/pull/2240#discussion_r475681222



##########
File path: hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ipc/DecayRpcScheduler.java
##########
@@ -483,6 +501,12 @@ private void recomputeScheduleCache() {
 
     for (Map.Entry<Object, List<AtomicLong>> entry : callCosts.entrySet()) {
       Object id = entry.getKey();
+      // The priority for service users is always 0
+      if (isServiceUser((String)id)) {

Review comment:
       Thanks for your comment, @Hexiaoqiao.
   I think you can adjust the existing FairCallQueue settings to deal with that case to a certain extent. You could consider increasing the weight of the highest priority queue, or extending the length of the highest priority queue. However, this simple feature is not a feature that can be used in all cases. If more fairness is needed among service users, then I believe that [HADOOP-15016](https://issues.apache.org/jira/browse/HADOOP-15016) would be required.
   
   The case where I think this feature would be useful is when you have service users who are regularly running critical ETL jobs. A user who can make huge RPC requests at a moment may not be suite for a service user.




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