[
https://issues.apache.org/jira/browse/YARN-4781?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16429269#comment-16429269
]
Sunil G commented on YARN-4781:
-------------------------------
Thanks [~eepayne].
{quote}I'm not seeing the {{FairOrderingPolicy}} consider pending + used
{quote}
I think below code also calls getMagnitude
{code:java}
protected class FairComparator implements Comparator<SchedulableEntity> {
@Override
public int compare(final SchedulableEntity r1, final SchedulableEntity r2) {
int res = (int) Math.signum( getMagnitude(r1) - getMagnitude(r2) );
return res;
}
}
private double getMagnitude(SchedulableEntity r) {
double mag = r.getSchedulingResourceUsage().getCachedUsed(
CommonNodeLabelsManager.ANY).getMemorySize();
if (sizeBasedWeight) {
double weight = Math.log1p(r.getSchedulingResourceUsage().getCachedDemand(
CommonNodeLabelsManager.ANY).getMemorySize()) / Math.log(2);
mag = mag / weight;
}
return mag;
}{code}
Also in {{AbstractComparatorOrderingPolicy}}
{code:java}
public static void updateSchedulingResourceUsage(ResourceUsage ru) {
ru.setCachedUsed(CommonNodeLabelsManager.ANY, ru.getAllUsed());
ru.setCachedPending(CommonNodeLabelsManager.ANY, ru.getAllPending());
}
{code}
Hence in case we use sizeBasedWeight, we are considering pending as well. So i
had this doubt..
bq.I don't think these comparators need to be {{Serializable}}, do they?
Yes. You are correct. I think, we dont need this.
> Support intra-queue preemption for fairness ordering policy.
> ------------------------------------------------------------
>
> Key: YARN-4781
> URL: https://issues.apache.org/jira/browse/YARN-4781
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: scheduler
> Reporter: Wangda Tan
> Assignee: Wangda Tan
> Priority: Major
> Attachments: YARN-4781.001.patch, YARN-4781.002.patch,
> YARN-4781.003.patch
>
>
> We introduced fairness queue policy since YARN-3319, which will let large
> applications make progresses and not starve small applications. However, if a
> large application takes the queue’s resources, and containers of the large
> app has long lifespan, small applications could still wait for resources for
> long time and SLAs cannot be guaranteed.
> Instead of wait for application release resources on their own, we need to
> preempt resources of queue with fairness policy enabled.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
