update dubhe-server/yaml/common-k8s.yaml.

dubhe-server/yaml/common-k8s.yaml

@@ -50,9 +50,9 @@ k8s:
   # 展示Pod的CPU使用率,Memory使用量,GPU使用率的grafana地址
   pod:
     metrics:
-      grafanaUrl: http://127.0.0.1:30006/d/job/monitor?orgId=1&refresh=5s&kiosk&var-pod=
+      grafanaUrl: http://grafana.dubhe.ai:30006/d/job/monitor?orgId=1&refresh=5s&kiosk&var-pod=
   prometheus:
-    url: http://127.0.0.1:30003/
+    url: http://10.5.26.91:30003/
     query: api/v1/query
     query-range: api/v1/query_range
     gpu-query-param: sum(container_accelerator_duty_cycle{pod="pod-name-placeholder"})by(pod,acc_id)
@@ -60,9 +60,16 @@ k8s:
     gpu-mem-use-query-param: sum(container_accelerator_memory_used_bytes{pod="pod-name-placeholder"})by(pod,acc_id)
     cpu-range-query-param: sum(rate(container_cpu_usage_seconds_total{image!="",pod="pod-name-placeholder"}[1m])) by (pod) / (sum(container_spec_cpu_quota{image!=""}/100000) by (pod)) * 100
     mem-range-query-param: sum(container_memory_rss{image!="",pod="pod-name-placeholder"})
+    gpu-usage-query-param: sum by(Hostname,gpu)(DCGM_FI_PROF_GR_ENGINE_ACTIVE{pod!=""})
     gpu-range-query-param: sum(container_accelerator_duty_cycle{pod="pod-name-placeholder"}) by (pod,acc_id)
     gpu-mem-total-range-query-param: sum(container_accelerator_memory_total_bytes{pod="pod-name-placeholder"}) by (pod,acc_id)
     gpu-mem-use-range-query-param: sum(container_accelerator_memory_used_bytes{pod="pod-name-placeholder"}) by (pod,acc_id)
+    gpu-usage-rate-query-param: topk(10,sort_desc(max_over_time(namespace:DCGM_FI_PROF_GR_ENGINE_ACTIVE:sumn[usage-rate-day])))
+    cpu-usage-rate-query-param: topk(10,sort_desc(max_over_time(namespace:container_cpu_user_seconds_total_sumn:raten[usage-rate-day])))
+    mem-usage-rate-query-param: topk(10,sort_desc(max_over_time(namespace:CONTAINER_MEMERY_USAGE_BYTES:sumn{namespace=~"namespace.*"}[usage-rate-day])))
+    gpu-usage-namespace-query-param: max_over_time(namespace:DCGM_FI_PROF_GR_ENGINE_ACTIVE:sumn{namespace=~"namespace-placeholder"}[usage-rate-day])
+    cpu-usage-namespace-query-param: max_over_time(namespace:container_cpu_user_seconds_total_sumn:raten{namespace=~"namespace-placeholder"}[usage-rate-day])
+    mem-usage-namespace-query-param: max_over_time(namespace:CONTAINER_MEMERY_USAGE_BYTES:sumn{namespace=~"namespace-placeholder"}[usage-rate-day])
   nfs-storage-class-name: zjlab-nfs-storage
   namespace-limits:
     cpu: 10