[ https://issues.apache.org/jira/browse/YARN-11633?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17796954#comment-17796954 ]
ASF GitHub Bot commented on YARN-11633: --------------------------------------- slfan1989 commented on PR #6356: URL: https://github.com/apache/hadoop/pull/6356#issuecomment-1857019443 @singer-bin Thank you forYARN's scheduler compares based on memory by default. We can refer to `FairSharePolicy.java` Therefore, using memory for comparison in LoadBasedRouterPolicy is in line with the design rules of the scheduler. ``` /** * Makes scheduling decisions by trying to equalize shares of memory. */ @Private @Unstable public class FairSharePolicy extends SchedulingPolicy { @VisibleForTesting public static final String NAME = "fair"; private static final Logger LOG = LoggerFactory.getLogger(FairSharePolicy.class); private static final String MEMORY = ResourceInformation.MEMORY_MB.getName(); private static final DefaultResourceCalculator RESOURCE_CALCULATOR = new DefaultResourceCalculator(); private static final FairShareComparator COMPARATOR = new FairShareComparator(); @Override public String getName() { return NAME; } ... ``` > [Federation] Improve LoadBasedRouterPolicy To Use Available vcores > ------------------------------------------------------------------ > > Key: YARN-11633 > URL: https://issues.apache.org/jira/browse/YARN-11633 > Project: Hadoop YARN > Issue Type: Improvement > Components: federation > Affects Versions: 3.3.6 > Reporter: yanbin.zhang > Priority: Major > Labels: pull-request-available > > When selecting a subcluster, consider not only available memory but also > available vcore -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org