[QIYI] feat(coordinator): add app running label metrics

zuston · Mar 19, 2024 · 8bb8c64 · 8bb8c64
1 parent 5044966
commit 8bb8c64
Show file tree

Hide file tree

Showing 2 changed files with 7 additions and 0 deletions.
diff --git a/coordinator/src/main/java/org/apache/uniffle/coordinator/ApplicationManager.java b/coordinator/src/main/java/org/apache/uniffle/coordinator/ApplicationManager.java
@@ -348,8 +348,12 @@ protected void statusCheck() {
         }
 
         CoordinatorMetrics.gaugeTaskFailedNum.remove(appId);
+        CoordinatorMetrics.gaugeAppRunning.remove(appId);
       }
       CoordinatorMetrics.gaugeRunningAppNum.set(appIds.size());
+      for (String appId : appIds.keySet()) {
+        CoordinatorMetrics.gaugeAppRunning.labels(appId).set(0);
+      }
       updateRemoteStorageMetrics();
       if (quotaManager != null) {
         quotaManager.updateQuotaMetrics();

diff --git a/coordinator/src/main/java/org/apache/uniffle/coordinator/metric/CoordinatorMetrics.java b/coordinator/src/main/java/org/apache/uniffle/coordinator/metric/CoordinatorMetrics.java
@@ -47,6 +47,8 @@ public class CoordinatorMetrics {
   public static final String USER_LABEL = "user_name";
 
   private static final String TASK_FAILED_NUM = "task_failed_num";
+  private static final String APP_RUNNING = "app_running";
+  public static Gauge gaugeAppRunning;
   public static Gauge gaugeTaskFailedNum;
   public static Gauge gaugeTotalServerNum;
   public static Gauge gaugeExcludeServerNum;
@@ -115,6 +117,7 @@ public static void updateDynamicGaugeForRemoteStorage(String storageHost, double
 
   private static void setUpMetrics() {
     gaugeTaskFailedNum = metricsManager.addGauge(TASK_FAILED_NUM, "app_id");
+    gaugeAppRunning = metricsManager.addGauge(APP_RUNNING, "app_id");
     gaugeTotalServerNum = metricsManager.addGauge(TOTAL_SERVER_NUM);
     gaugeExcludeServerNum = metricsManager.addGauge(EXCLUDE_SERVER_NUM);
     gaugeUnhealthyServerNum = metricsManager.addGauge(UNHEALTHY_SERVER_NUM);