YARN-10713. ClusterMetrics should support custom resource capacity related metrics. Contributed by Qi Zhu.

Author: ericbadger

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ClusterMetrics.java

@@ -20,6 +20,7 @@
 
 import static org.apache.hadoop.metrics2.lib.Interns.info;
 
+import java.util.Map;
 import java.util.concurrent.atomic.AtomicBoolean;
 
 import org.apache.hadoop.classification.InterfaceAudience;
@@ -35,6 +36,9 @@
 import org.apache.hadoop.yarn.api.records.Resource;
 import org.apache.hadoop.thirdparty.com.google.common.annotations.VisibleForTesting;
 import org.apache.hadoop.yarn.api.records.ResourceInformation;
+import org.apache.hadoop.yarn.metrics.CustomResourceMetricValue;
+import org.apache.hadoop.yarn.metrics.CustomResourceMetrics;
+import org.apache.hadoop.yarn.server.resourcemanager.scheduler.QueueMetricsForCustomResources;
 import org.apache.hadoop.yarn.util.resource.ResourceUtils;
 
 @InterfaceAudience.Private
@@ -58,10 +62,19 @@ public class ClusterMetrics {
   @Metric("Vcore Utilization") MutableGaugeLong utilizedVirtualCores;
   @Metric("Memory Capability") MutableGaugeLong capabilityMB;
   @Metric("Vcore Capability") MutableGaugeLong capabilityVirtualCores;
-  @Metric("GPU Capability") MutableGaugeLong capabilityGPUs;
 
   private static final MetricsInfo RECORD_INFO = info("ClusterMetrics",
   "Metrics for the Yarn Cluster");
+
+  private static final String CUSTOM_RESOURCE_CAPABILITY_METRIC_PREFIX =
+      "Capability.";
+  private static final String CUSTOM_RESOURCE_CAPABILITY_METRIC_DESC =
+      "NAME Capability";
+
+  private static CustomResourceMetrics customResourceMetrics;
+
+  private final CustomResourceMetricValue customResourceCapability =
+      new CustomResourceMetricValue();
 
   private static volatile ClusterMetrics INSTANCE = null;
   private static MetricsRegistry registry;
@@ -86,6 +99,17 @@ private static void registerMetrics() {
     if (ms != null) {
       ms.register("ClusterMetrics", "Metrics for the Yarn Cluster", INSTANCE);
     }
+
+    if (ResourceUtils.getNumberOfKnownResourceTypes() > 2) {
+      customResourceMetrics =
+          new CustomResourceMetrics();
+      Map<String, Long> customResources =
+          customResourceMetrics.initAndGetCustomResources();
+      customResourceMetrics.
+          registerCustomResources(customResources,
+              registry, CUSTOM_RESOURCE_CAPABILITY_METRIC_PREFIX,
+              CUSTOM_RESOURCE_CAPABILITY_METRIC_DESC);
+    }
   }
 
   @VisibleForTesting
@@ -209,23 +233,20 @@ public long getCapabilityVirtualCores() {
     return capabilityVirtualCores.value();
   }
 
-  public long getCapabilityGPUs() {
-    if (capabilityGPUs == null) {
-      return 0;
-    }
+  public Map<String, Long> getCustomResourceCapability() {
+    return customResourceCapability.getValues();
+  }
 
-    return capabilityGPUs.value();
+  public void setCustomResourceCapability(Resource res) {
+    this.customResourceCapability.set(res);
   }
 
   public void incrCapability(Resource res) {
     if (res != null) {
       capabilityMB.incr(res.getMemorySize());
       capabilityVirtualCores.incr(res.getVirtualCores());
-      Integer gpuIndex = ResourceUtils.getResourceTypeIndex()
-          .get(ResourceInformation.GPU_URI);
-      if (gpuIndex != null) {
-        capabilityGPUs.incr(res.
-            getResourceValue(ResourceInformation.GPU_URI));
+      if (customResourceCapability != null) {
+        customResourceCapability.increase(res);
       }
     }
   }
@@ -234,11 +255,8 @@ public void decrCapability(Resource res) {
     if (res != null) {
       capabilityMB.decr(res.getMemorySize());
       capabilityVirtualCores.decr(res.getVirtualCores());
-      Integer gpuIndex = ResourceUtils.getResourceTypeIndex()
-          .get(ResourceInformation.GPU_URI);
-      if (gpuIndex != null) {
-        capabilityGPUs.decr(res.
-            getResourceValue(ResourceInformation.GPU_URI));
+      if (customResourceCapability != null) {
+        customResourceCapability.decrease(res);
       }
     }
   }

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/TestCSAllocateCustomResource.java

@@ -231,7 +231,8 @@ public void testClusterMetricsWithGPU()
     assertEquals("Cluster Capability Vcores incorrect",
         metrics.getCapabilityVirtualCores(), 4 * 8);
     assertEquals("Cluster Capability GPUs incorrect",
-        metrics.getCapabilityGPUs(), 4 * 8);
+        (metrics.getCustomResourceCapability()
+            .get(GPU_URI)).longValue(), 4 * 8);
 
     for (RMNode rmNode : rmNodes) {
       nodeTracker.removeNode(rmNode.getNodeID());
@@ -243,7 +244,8 @@ public void testClusterMetricsWithGPU()
     assertEquals("Cluster Capability Vcores incorrect",
         metrics.getCapabilityVirtualCores(), 0);
     assertEquals("Cluster Capability GPUs incorrect",
-        metrics.getCapabilityGPUs(), 0);
+        (metrics.getCustomResourceCapability()
+            .get(GPU_URI)).longValue(), 0);
     ClusterMetrics.destroy();
   }
 }