Merge pull request #25678 from vespa-engine/yngveaasheim/use-enums-for-cluster-controller-metrics

Use enums for cluster controller metrics

4 files changed, 60 insertions, 29 deletions

diff --git a/config-model/src/main/java/com/yahoo/vespa/model/admin/monitoring/VespaMetricSet.java b/config-model/src/main/java/com/yahoo/vespa/model/admin/monitoring/VespaMetricSet.java
index 7d4c4129640..d73385831ff 100644
--- a/config-model/src/main/java/com/yahoo/vespa/model/admin/monitoring/VespaMetricSet.java
+++ b/config-model/src/main/java/com/yahoo/vespa/model/admin/monitoring/VespaMetricSet.java
@@ -237,32 +237,32 @@ public class VespaMetricSet {
     private static Set<Metric> getClusterControllerMetrics() {
         Set<Metric> metrics = new LinkedHashSet<>();
 
-        metrics.add(new Metric("cluster-controller.down.count.last"));
-        metrics.add(new Metric("cluster-controller.initializing.count.last"));
-        metrics.add(new Metric("cluster-controller.maintenance.count.last"));
-        metrics.add(new Metric("cluster-controller.retired.count.last"));
-        metrics.add(new Metric("cluster-controller.stopping.count.last"));
-        metrics.add(new Metric("cluster-controller.up.count.last"));
-        metrics.add(new Metric("cluster-controller.cluster-state-change.count"));
-        addMetric(metrics, "cluster-controller.busy-tick-time-ms", List.of("last", "max", "sum", "count"));
-        addMetric(metrics, "cluster-controller.idle-tick-time-ms", List.of("last", "max", "sum", "count"));
-
-        addMetric(metrics, "cluster-controller.work-ms", List.of("last", "sum", "count"));
-
-        metrics.add(new Metric("cluster-controller.is-master.last"));
-        metrics.add(new Metric("cluster-controller.remote-task-queue.size.last"));
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_DOWN_COUNT.last());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_INITIALIZING_COUNT.last());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_MAINTENANCE_COUNT.last());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_RETIRED_COUNT.last());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_STOPPING_COUNT.last());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_UP_COUNT.last());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_CLUSTER_STATE_CHANGE_COUNT.baseName());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_BUSY_TICK_TIME_MS, EnumSet.of(last, max, sum, count));
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_IDLE_TICK_TIME_MS, EnumSet.of(last, max, sum, count));
+        
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_WORK_MS, EnumSet.of(last, sum, count));
+        
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_IS_MASTER.last());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_REMOTE_TASK_QUEUE_SIZE.last());
         // TODO(hakonhall): Update this name once persistent "count" metrics has been implemented.
         // DO NOT RELY ON THIS METRIC YET.
-        metrics.add(new Metric("cluster-controller.node-event.count"));
-
-        addMetric(metrics, "cluster-controller.resource_usage.nodes_above_limit", List.of("last", "max"));
-        addMetric(metrics, "cluster-controller.resource_usage.max_memory_utilization", List.of("last", "max"));
-        addMetric(metrics, "cluster-controller.resource_usage.max_disk_utilization", List.of("last", "max"));
-        metrics.add(new Metric("cluster-controller.resource_usage.disk_limit.last"));
-        metrics.add(new Metric("cluster-controller.resource_usage.memory_limit.last"));
-
-        metrics.add(new Metric("reindexing.progress.last"));
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_NODE_EVENT_COUNT.baseName());
+        
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_NODES_ABOVE_LIMIT, EnumSet.of(last, max));
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_MAX_MEMORY_UTILIZATION, EnumSet.of(last, max));
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_MAX_DISK_UTILIZATION, EnumSet.of(last, max));
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_MEMORY_LIMIT.last());
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_DISK_LIMIT.last());
 
+        addMetric(metrics, ContainerMetrics.CLUSTER_CONTROLLER_REINDEXING_PROGRESS.last());
+        
         return metrics;
     }
 
diff --git a/configserver/src/main/java/com/yahoo/vespa/config/server/metrics/ClusterDeploymentMetricsRetriever.java b/configserver/src/main/java/com/yahoo/vespa/config/server/metrics/ClusterDeploymentMetricsRetriever.java
index c1bc1027690..98f01bada3f 100644
--- a/configserver/src/main/java/com/yahoo/vespa/config/server/metrics/ClusterDeploymentMetricsRetriever.java
+++ b/configserver/src/main/java/com/yahoo/vespa/config/server/metrics/ClusterDeploymentMetricsRetriever.java
@@ -3,6 +3,7 @@ package com.yahoo.vespa.config.server.metrics;
 
 import ai.vespa.util.http.hc5.VespaHttpClientBuilder;
 import com.yahoo.concurrent.DaemonThreadFactory;
+import com.yahoo.metrics.ContainerMetrics;
 import com.yahoo.slime.ArrayTraverser;
 import com.yahoo.slime.Cursor;
 import com.yahoo.slime.Inspector;
@@ -145,12 +146,12 @@ public class ClusterDeploymentMetricsRetriever {
                         .ifPresent(docCount -> aggregator.get().addDocumentCount(docCount));
                 break;
             case VESPA_CONTAINER_CLUSTERCONTROLLER:
-                optionalDouble(values.field("cluster-controller.resource_usage.max_memory_utilization.max")).ifPresent(memoryUtil ->
+                optionalDouble(values.field(ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_MAX_MEMORY_UTILIZATION.max())).ifPresent(memoryUtil ->
                         aggregator.get()
                                 .addMemoryUsage(memoryUtil,
-                                        values.field("cluster-controller.resource_usage.memory_limit.last").asDouble())
-                                .addDiskUsage(values.field("cluster-controller.resource_usage.max_disk_utilization.max").asDouble(),
-                                        values.field("cluster-controller.resource_usage.disk_limit.last").asDouble()));
+                                        values.field(ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_MEMORY_LIMIT.last()).asDouble())
+                                .addDiskUsage(values.field(ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_MAX_DISK_UTILIZATION.max()).asDouble(),
+                                        values.field(ContainerMetrics.CLUSTER_CONTROLLER_RESOURCE_USAGE_DISK_LIMIT.last()).asDouble()));
                 break;
         }
     }
diff --git a/container-core/src/main/java/com/yahoo/metrics/ContainerMetrics.java b/container-core/src/main/java/com/yahoo/metrics/ContainerMetrics.java
index 513cf499362..35feb291c84 100644
--- a/container-core/src/main/java/com/yahoo/metrics/ContainerMetrics.java
+++ b/container-core/src/main/java/com/yahoo/metrics/ContainerMetrics.java
@@ -150,8 +150,34 @@ public enum ContainerMetrics {
     // Deprecated metrics. TODO: Remove on Vespa 9.
     SERVER_REJECTED_REQUESTS("serverRejectedRequests", Unit.OPERATION, "Deprecated. Use jdisc.thread_pool.rejected_tasks instead."),
     SERVER_THREAD_POOL_SIZE("serverThreadPoolSize", Unit.THREAD, "Deprecated. Use jdisc.thread_pool.size instead."),
-    SERVER_ACTIVE_THREADS("serverActiveThreads", Unit.THREAD, "Deprecated. Use jdisc.thread_pool.active_threads instead.");
-        
+    SERVER_ACTIVE_THREADS("serverActiveThreads", Unit.THREAD, "Deprecated. Use jdisc.thread_pool.active_threads instead."),
+
+    
+    // Metrics from the cluster controller
+    CLUSTER_CONTROLLER_DOWN_COUNT("cluster-controller.down.count", Unit.NODE, "Number of content nodes down"),
+    CLUSTER_CONTROLLER_INITIALIZING_COUNT("cluster-controller.initializing.count", Unit.NODE, "Number of content nodes initializing"),
+    CLUSTER_CONTROLLER_MAINTENANCE_COUNT("cluster-controller.maintenance.count", Unit.NODE, "Number of content nodes in maintenance"),
+    CLUSTER_CONTROLLER_RETIRED_COUNT("cluster-controller.retired.count", Unit.NODE, "Number of content nodes that are retired"),
+    CLUSTER_CONTROLLER_STOPPING_COUNT("cluster-controller.stopping.count", Unit.NODE, "Number of content nodes currently stopping"),
+    CLUSTER_CONTROLLER_UP_COUNT("cluster-controller.up.count", Unit.NODE, "Number of content nodes up"),
+    CLUSTER_CONTROLLER_CLUSTER_STATE_CHANGE_COUNT("cluster-controller.cluster-state-change.count", Unit.NODE, "Number of nodes changing state"),
+    CLUSTER_CONTROLLER_BUSY_TICK_TIME_MS("cluster-controller.busy-tick-time-ms", Unit.MILLISECOND, "Time busy"),
+    CLUSTER_CONTROLLER_IDLE_TICK_TIME_MS("cluster-controller.idle-tick-time-ms", Unit.MILLISECOND, "Time idle"),
+    CLUSTER_CONTROLLER_WORK_MS("cluster-controller.work-ms", Unit.MILLISECOND, "Time used for actual work"),
+    CLUSTER_CONTROLLER_IS_MASTER("cluster-controller.is-master", Unit.BINARY, "1 if this cluster controller is currently the master, or 0 if not"),
+    CLUSTER_CONTROLLER_REMOTE_TASK_QUEUE_SIZE("cluster-controller.remote-task-queue.size", Unit.OPERATION, "Number of remote tasks queued"),
+    // TODO(hakonhall): Update this name once persistent "count" metrics has been implemented.
+    // DO NOT RELY ON THIS METRIC YET.
+    CLUSTER_CONTROLLER_NODE_EVENT_COUNT("cluster-controller.node-event.count", Unit.OPERATION, "Number of node events"),
+    CLUSTER_CONTROLLER_RESOURCE_USAGE_NODES_ABOVE_LIMIT("cluster-controller.resource_usage.nodes_above_limit", Unit.NODE, "The number of content nodes above resource limit, blocking feed"),
+    CLUSTER_CONTROLLER_RESOURCE_USAGE_MAX_MEMORY_UTILIZATION("cluster-controller.resource_usage.max_memory_utilization", Unit.FRACTION, "Current memory utilisation, per content node"),
+    CLUSTER_CONTROLLER_RESOURCE_USAGE_MAX_DISK_UTILIZATION("cluster-controller.resource_usage.max_disk_utilization", Unit.FRACTION, "Current disk space utilisation, per content node"),
+    CLUSTER_CONTROLLER_RESOURCE_USAGE_MEMORY_LIMIT("cluster-controller.resource_usage.memory_limit", Unit.FRACTION, "Disk space limit as a fraction of available disk space"),
+    CLUSTER_CONTROLLER_RESOURCE_USAGE_DISK_LIMIT("cluster-controller.resource_usage.disk_limit", Unit.FRACTION, "Memory space limit as a fraction of available memory"),
+    CLUSTER_CONTROLLER_REINDEXING_PROGRESS("reindexing.progress", Unit.FRACTION, "Re-indexing progress");
+
+    
+    
     private final String name;
     private final Unit unit;
     private final String description;
diff --git a/container-core/src/main/java/com/yahoo/metrics/Unit.java b/container-core/src/main/java/com/yahoo/metrics/Unit.java
index 5a4a0108534..58289143337 100644
--- a/container-core/src/main/java/com/yahoo/metrics/Unit.java
+++ b/container-core/src/main/java/com/yahoo/metrics/Unit.java
@@ -5,6 +5,7 @@ package com.yahoo.metrics;
  */
 public enum Unit {
 
+    BINARY(BaseUnit.BINARY),
     BYTE(BaseUnit.BYTE),
     CONNECTION(BaseUnit.CONNECTION),
     DOCUMENT(BaseUnit.DOCUMENT),
@@ -14,6 +15,7 @@ public enum Unit {
     HIT_PER_QUERY(BaseUnit.HIT, BaseUnit.QUERY),
     ITEM(BaseUnit.ITEM),
     MILLISECOND(BaseUnit.MILLISECOND),
+    NODE(BaseUnit.NODE),
     OPERATION(BaseUnit.OPERATION),
     OPERATION_PER_SECOND(BaseUnit.OPERATION, BaseUnit.SECOND),
     QUERY(BaseUnit.QUERY),
@@ -53,6 +55,7 @@ public enum Unit {
 
     private enum BaseUnit {
 
+        BINARY("binary"),
         BYTE("byte"),
         CONNECTION("connection"),
         DOCUMENT("document"),
@@ -60,6 +63,7 @@ public enum Unit {
         HIT("hit"),
         ITEM("item"),
         MILLISECOND("millisecond", "ms"),
+        NODE("node"),
         OPERATION("operation"),
         QUERY("query"),
         REQUEST("request"),