Skip to content

YARN-11633.[Federation] Improve LoadBasedRouterPolicy To Use Available vcores. #6356

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

YARN-11633.[Federation] Improve LoadBasedRouterPolicy To Use Available vcores. #6356

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 19 additions & 1 deletion .../java/org/apache/hadoop/yarn/server/federation/policies/router/LoadBasedRouterPolicy.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,12 +66,19 @@ protected SubClusterId chooseSubCluster(
Map<SubClusterIdInfo, Float> weights = getPolicyInfo().getRouterPolicyWeights();
SubClusterIdInfo chosen = null;
long currBestMem = -1;
long currBestVcore = -1;
for (Map.Entry<SubClusterId, SubClusterInfo> entry : preSelectSubclusters.entrySet()) {
SubClusterIdInfo id = new SubClusterIdInfo(entry.getKey());
if (weights.containsKey(id) && weights.get(id) > 0) {
long availableMemory = getAvailableMemory(entry.getValue());
if (availableMemory > currBestMem) {
long availableVcore = getAvailableVcore(entry.getValue());
if (availableMemory > currBestMem && availableVcore > currBestVcore) {
Copy link
Contributor

@slfan1989 slfan1989 Dec 15, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more. 

if (availableMemory > currBestMem) {
 ....
}

This code is needed to select the largest subcluster of availableMemory. Memory is a more strictly limited resource, because without memory, the application cannot execute. Sufficient memory can also be considered as the cluster is idle.

If we add cpu constraints (availableVcore), subclusters with smaller memory may be selected, which is not an expected situation.

currBestMem = availableMemory;
currBestVcore = availableVcore;
chosen = id;
} else if (availableMemory > currBestMem) {
currBestMem = availableMemory;
currBestVcore = availableVcore;
chosen = id;
}
}
Expand All @@ -93,4 +100,15 @@ private long getAvailableMemory(SubClusterInfo value) throws YarnException {
throw new YarnException("FederationSubClusterInfo cannot be parsed", j);
}
}

private long getAvailableVcore(SubClusterInfo value) throws YarnException {
try {
long vcore = -1;
JSONObject obj = new JSONObject(value.getCapability());
vcore = obj.getJSONObject("clusterMetrics").getLong("availableVirtualCores");
return vcore;
} catch (JSONException j) {
throw new YarnException("FederationSubClusterInfo cannot be parsed", j);
}
}
}