XComp commented on code in PR #25218:
URL: https://github.com/apache/flink/pull/25218#discussion_r2085045943


##########
flink-core/src/main/java/org/apache/flink/configuration/JobManagerOptions.java:
##########
@@ -677,6 +679,31 @@ public InlineElement getDescription() {
                                             
code(SchedulerType.AdaptiveBatch.name()))
                                     .build());
 
+    @Experimental
+    @Documentation.Section({
+        Documentation.Sections.EXPERT_SCHEDULING,
+        Documentation.Sections.ALL_JOB_MANAGER
+    })
+    public static final ConfigOption<Boolean> 
SCHEDULER_PREFER_MINIMAL_TASKMANAGERS_ENABLED =
+            key("jobmanager.adaptive-scheduler.prefer-minimal-taskmanagers")
+                    .booleanType()
+                    .defaultValue(true)

Review Comment:
   > After the mailing list and PR conversation IMO the consensus is that 
minimizing TM number is desired in most use-cases, so I am leaning towards 
changing this option even though it's a patch release, and include some 
warning/details about it in the Release Notes.
   > 
   > I would say most people do not read docs only if something do not work as 
they think it should, so I would prefer making it work as one would expect by 
default the most as we can. But I do not necessarily insist, if we decide to 
opt this out, we can still highlight to opt-in in the release notes, but back 
to base, most people won't read that anyways. :)
   
   So, if it doesn't work as expected (i.e. that too many TMs being used) they 
should be eventually guided to FLINK-33977, this PR or the configuration 
parameter. And we could also highlight that in the release notes.
   
   But I also see your point that the ML discussion considered it a "some-what 
bug"/unexpected behavior.
   
   For me, introducing this feature with the feature flag/config parameter 
being disabled is still the best compromise between version stability and 
fixing this bug.



-- 
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: issues-unsubscr...@flink.apache.org

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

Reply via email to