Aggregate in case of node change

robusta-dev · Sep 4, 2023 · a9d281d · a9d281d
1 parent 3d0b3bc
commit a9d281d
Show file tree

Hide file tree

Showing 3 changed files with 49 additions and 35 deletions.
diff --git a/robusta_krr/core/integrations/prometheus/metrics/cpu.py b/robusta_krr/core/integrations/prometheus/metrics/cpu.py
@@ -8,14 +8,16 @@ def get_query(self, object: K8sObjectData, duration: str, step: str) -> str:
         pods_selector = "|".join(pod.name for pod in object.pods)
         cluster_label = self.get_prometheus_cluster_label()
         return f"""
-            rate(
-                container_cpu_usage_seconds_total{{
-                    namespace="{object.namespace}",
-                    pod=~"{pods_selector}",
-                    container="{object.container}"
-                    {cluster_label}
-                }}[{step}]
-            )
+            sum(
+                rate(
+                    container_cpu_usage_seconds_total{{
+                        namespace="{object.namespace}",
+                        pod=~"{pods_selector}",
+                        container="{object.container}"
+                        {cluster_label}
+                    }}[{step}]
+                ) 
+            ) by (container, pod, job)
         """
 
 
@@ -27,14 +29,17 @@ def get_query(self, object: K8sObjectData, duration: str, step: str) -> str:
             return f"""
                 quantile_over_time(
                     {round(percentile / 100, 2)},
-                    rate(
-                        container_cpu_usage_seconds_total{{
-                            namespace="{object.namespace}",
-                            pod=~"{pods_selector}",
-                            container="{object.container}"
-                            {cluster_label}
-                        }}[{step}]
-                    )[{duration}:{step}]
+                    sum(
+                        rate(
+                            container_cpu_usage_seconds_total{{
+                                namespace="{object.namespace}",
+                                pod=~"{pods_selector}",
+                                container="{object.container}"
+                                {cluster_label}
+                            }}[{step}]
+                        )
+                    ) by (container, pod, job)
+                    [{duration}:{step}]
                 )
             """
 
@@ -47,11 +52,14 @@ def get_query(self, object: K8sObjectData, duration: str, step: str) -> str:
         cluster_label = self.get_prometheus_cluster_label()
         return f"""
             count_over_time(
-                container_cpu_usage_seconds_total{{
-                    namespace="{object.namespace}",
-                    pod=~"{pods_selector}",
-                    container="{object.container}"
-                    {cluster_label}
-                }}[{duration}]
+                sum(
+                    container_cpu_usage_seconds_total{{
+                        namespace="{object.namespace}",
+                        pod=~"{pods_selector}",
+                        container="{object.container}"
+                        {cluster_label}
+                    }}
+                ) by (container, pod, job)
+                [{duration}:{step}]
             )
         """
diff --git a/robusta_krr/core/integrations/prometheus/metrics/memory.py b/robusta_krr/core/integrations/prometheus/metrics/memory.py
@@ -25,12 +25,15 @@ def get_query(self, object: K8sObjectData, duration: str, step: str) -> str:
         cluster_label = self.get_prometheus_cluster_label()
         return f"""
             max_over_time(
-                container_memory_working_set_bytes{{
-                    namespace="{object.namespace}",
-                    pod=~"{pods_selector}",
-                    container="{object.container}"
-                    {cluster_label}
-                }}[{duration}:{step}]
+                sum(
+                    container_memory_working_set_bytes{{
+                        namespace="{object.namespace}",
+                        pod=~"{pods_selector}",
+                        container="{object.container}"
+                        {cluster_label}
+                    }}
+                ) by (container, pod, job)
+                [{duration}:{step}]
             )
         """
 
@@ -41,11 +44,14 @@ def get_query(self, object: K8sObjectData, duration: str, step: str) -> str:
         cluster_label = self.get_prometheus_cluster_label()
         return f"""
             count_over_time(
-                container_memory_working_set_bytes{{
-                    namespace="{object.namespace}",
-                    pod=~"{pods_selector}",
-                    container="{object.container}"
-                    {cluster_label}
-                }}[{duration}:{step}]
+                sum(
+                    container_memory_working_set_bytes{{
+                        namespace="{object.namespace}",
+                        pod=~"{pods_selector}",
+                        container="{object.container}"
+                        {cluster_label}
+                    }}
+                ) by (container, pod, job)
+                [{duration}:{step}]
             )
         """
diff --git a/robusta_krr/strategies/simple.py b/robusta_krr/strategies/simple.py
@@ -88,7 +88,7 @@ def __calculate_cpu_proposal(
         if object_data.hpa is not None and object_data.hpa.target_cpu_utilization_percentage is not None:
             return ResourceRecommendation.undefined(info="HPA detected")
 
-        cpu_usage = self.settings.calculate_cpu_proposal(data)
+        cpu_usage = self.settings.calculate_cpu_proposal(filtered_data)
         return ResourceRecommendation(request=cpu_usage, limit=None)
 
     def __calculate_memory_proposal(