Merge pull request #31539 from vespa-engine/bjormel/peak_load_metrics

vespa-engine · Jun 11, 2024 · e5be68b · e5be68b
2 parents 26c753d + e33ac17
commit e5be68b
Show file tree

Hide file tree

Showing 3 changed files with 9 additions and 0 deletions.
diff --git a/metrics/src/main/java/ai/vespa/metrics/ConfigServerMetrics.java b/metrics/src/main/java/ai/vespa/metrics/ConfigServerMetrics.java
@@ -66,6 +66,9 @@ public enum ConfigServerMetrics implements VespaMetrics {
     CLUSTER_LOAD_IDEAL_CPU("cluster.load.ideal.cpu", Unit.FRACTION, "The ideal cpu load of a certain cluster"),
     CLUSTER_LOAD_IDEAL_MEMORY("cluster.load.ideal.memory", Unit.FRACTION, "The ideal memory load of a certain cluster"),
     CLUSTER_LOAD_IDEAL_DISK("cluster.load.ideal.disk", Unit.FRACTION, "The ideal disk load of a certain cluster"),
+    CLUSTER_LOAD_PEAK_CPU("cluster.load.peak.cpu", Unit.FRACTION, "The peak cpu load in the period considered of a certain cluster"),
+    CLUSTER_LOAD_PEAK_MEMORY("cluster.load.peak.memory", Unit.FRACTION, "The peak memory load in the period considered of a certain cluster"),
+    CLUSTER_LOAD_PEAK_DISK("cluster.load.peak.disk", Unit.FRACTION, "The peak disk load in the period considered of a certain cluster"),
 
     ZONE_WORKING("zone.working", Unit.BINARY, "The value 1 if zone is considered healthy, 0 if not. This is decided by considering the number of non-active nodes vs the number of active nodes in a zone"),
     CACHE_NODE_OBJECT_HIT_RATE("cache.nodeObject.hitRate", Unit.FRACTION, "The fraction of cache hits vs cache lookups for the node object cache"),

diff --git a/metrics/src/main/java/ai/vespa/metrics/set/InfrastructureMetricSet.java b/metrics/src/main/java/ai/vespa/metrics/set/InfrastructureMetricSet.java
@@ -67,6 +67,9 @@ private static Set<Metric> getConfigServerMetrics() {
         addMetric(metrics, ConfigServerMetrics.CLUSTER_LOAD_IDEAL_CPU.max());
         addMetric(metrics, ConfigServerMetrics.CLUSTER_LOAD_IDEAL_MEMORY.max());
         addMetric(metrics, ConfigServerMetrics.CLUSTER_LOAD_IDEAL_DISK.max());
+        addMetric(metrics, ConfigServerMetrics.CLUSTER_LOAD_PEAK_CPU.max());
+        addMetric(metrics, ConfigServerMetrics.CLUSTER_LOAD_PEAK_MEMORY.max());
+        addMetric(metrics, ConfigServerMetrics.CLUSTER_LOAD_PEAK_DISK.max());
         addMetric(metrics, ConfigServerMetrics.NODES_EMPTY_EXCLUSIVE.max());
         addMetric(metrics, ConfigServerMetrics.NODES_EXPIRED_DEPROVISIONED.count());
         addMetric(metrics, ConfigServerMetrics.NODES_EXPIRED_DIRTY.count());

diff --git a/...epository/src/main/java/com/yahoo/vespa/hosted/provision/maintenance/MetricsReporter.java b/...epository/src/main/java/com/yahoo/vespa/hosted/provision/maintenance/MetricsReporter.java
@@ -143,6 +143,9 @@ private void updateClusterCostMetrics(ClusterId clusterId,
         metric.set(ConfigServerMetrics.CLUSTER_LOAD_IDEAL_CPU.baseName(), cluster.get().target().ideal().cpu(), context);
         metric.set(ConfigServerMetrics.CLUSTER_LOAD_IDEAL_MEMORY.baseName(), cluster.get().target().ideal().memory(), context);
         metric.set(ConfigServerMetrics.CLUSTER_LOAD_IDEAL_DISK.baseName(), cluster.get().target().ideal().disk(), context);
+        metric.set(ConfigServerMetrics.CLUSTER_LOAD_PEAK_CPU.baseName(), cluster.get().target().peak().cpu(), context);
+        metric.set(ConfigServerMetrics.CLUSTER_LOAD_PEAK_MEMORY.baseName(), cluster.get().target().peak().memory(), context);
+        metric.set(ConfigServerMetrics.CLUSTER_LOAD_PEAK_DISK.baseName(), cluster.get().target().peak().disk(), context);
     }
 
     private void updateZoneMetrics() {