chore: rewrite queries for nim new graphs

Signed-off-by: Tomer Figenblat <[email protected]>
opendatahub-io · Dec 20, 2024 · 7785941 · 7785941
1 parent 9b9a528
commit 7785941
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 9 deletions.
diff --git a/controllers/constants/constants.go b/controllers/constants/constants.go
@@ -70,6 +70,7 @@ const (
 	DefaultStorageConfig             = "storage-config"
 	IntervalValue                    = "1m"
 	RequestRateInterval              = "5m"
+	GPUKVCacheSamplingInterval       = "24h"
 	OvmsImageName                    = "openvino_model_server"
 	TgisImageName                    = "text-generation-inference"
 	VllmImageName                    = "vllm"

diff --git a/controllers/constants/runtime-metrics.go b/controllers/constants/runtime-metrics.go
@@ -247,7 +247,7 @@ const (
 				"queries": [
 					{
 						"title": "Average e2e latency",
-						"query": "sum by (model_name) (rate(e2e_request_latency_seconds_sum{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}]) * 1000) / sum by (model_name) (rate(e2e_request_latency_seconds_count{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}]) * 1000)"
+						"query": "(rate(e2e_request_latency_seconds_sum{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}]) * 1000) / (rate(e2e_request_latency_seconds_count{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}]) * 1000)"
 					}
 				]
 			},
@@ -277,7 +277,7 @@ const (
 				"queries": [
 					{
 						"title": "GPU cache usage over time",
-						"query":  "round(sum(increase(gpu_cache_usage_perc{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}])))"
+						"query": "sum_over_time(gpu_cache_usage_perc{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${KV_CACHE_SAMPLING_RATE}])"
 					}
 				]
 			},
@@ -305,11 +305,11 @@ const (
 				"queries": [
 					{
 						"title": "Total prompts token",
-						"query":  "round(sum(increase(prompt_tokens_total{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}])))"
+						"query":  "round(rate(prompt_tokens_total{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}]))"
 					},
 					{
 						"title": "Total generation token",
-						"query":  "round(sum(increase(generation_tokens_total{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}])))"
+						"query":  "round(rate(generation_tokens_total{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}]))"
 					}
 				]
 			},
@@ -319,7 +319,7 @@ const (
 				"queries": [
 					{
 						"title": "Time to first token",
-						"query":  "rate(time_to_first_token_seconds_bucket{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}])"
+						"query": "rate(time_to_first_token_seconds_sum{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}])"
 					}
 				]
 			},
@@ -329,7 +329,7 @@ const (
 				"queries": [
 					{
 						"title": "Time per output token",
-						"query":  "rate(time_per_output_token_seconds_bucket{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}])"
+						"query": "rate(time_per_output_token_seconds_sum{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${RATE_INTERVAL}])"
 					}
 				]
 			},
@@ -339,11 +339,11 @@ const (
 				"queries": [
 					{
 						"title": "Number of successful incoming requests",
-						"query": "request_success_total{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}"
+						"query": "round(sum(increase(request_success_total{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${REQUEST_RATE_INTERVAL}])))"
 					},
 					{
 						"title": "Number of failed incoming requests",
-						"query": "request_failure_total{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}"
+						"query": "round(sum(increase(request_failure_total{namespace='${NAMESPACE}', pod=~'${MODEL_NAME}-predictor-.*'}[${REQUEST_RATE_INTERVAL}])))"
 					}
 				]
 			}

diff --git a/controllers/utils/utils.go b/controllers/utils/utils.go
@@ -427,6 +427,7 @@ func SubstituteVariablesInQueries(data string, namespace string, name string) st
 		"${NAMESPACE}", namespace,
 		"${MODEL_NAME}", name,
 		"${RATE_INTERVAL}", constants.IntervalValue,
-		"${REQUEST_RATE_INTERVAL}", constants.RequestRateInterval)
+		"${REQUEST_RATE_INTERVAL}", constants.RequestRateInterval,
+		"${KV_CACHE_SAMPLING_RATE}", constants.GPUKVCacheSamplingInterval)
 	return replacer.Replace(data)
 }