Add constants and skip metrics based on constant

Signed-off-by: Trey Dockendorf <[email protected]>

bindings/go/dcgm/const.go

@@ -14,11 +14,27 @@ type FieldValue_v1 struct {
 }
 
 const (
-	DCGM_FT_BINARY    = uint('b')
-	DCGM_FT_DOUBLE    = uint('d')
-	DCGM_FT_INT64     = uint('i')
-	DCGM_FT_STRING    = uint('s')
-	DCGM_FT_TIMESTAMP = uint('t')
+	DCGM_FT_BINARY                 = uint('b')
+	DCGM_FT_DOUBLE                 = uint('d')
+	DCGM_FT_INT64                  = uint('i')
+	DCGM_FT_STRING                 = uint('s')
+	DCGM_FT_TIMESTAMP              = uint('t')
+	DCGM_FT_INT32_BLANK            = int64(2147483632)
+	DCGM_FT_INT32_NOT_FOUND        = int64(DCGM_FT_INT32_BLANK + 1)
+	DCGM_FT_INT32_NOT_SUPPORTED    = int64(DCGM_FT_INT32_BLANK + 2)
+	DCGM_FT_INT32_NOT_PERMISSIONED = int64(DCGM_FT_INT32_BLANK + 3)
+	DCGM_FT_INT64_BLANK            = int64(9223372036854775792)
+	DCGM_FT_INT64_NOT_FOUND        = int64(DCGM_FT_INT64_BLANK + 1)
+	DCGM_FT_INT64_NOT_SUPPORTED    = int64(DCGM_FT_INT64_BLANK + 2)
+	DCGM_FT_INT64_NOT_PERMISSIONED = int64(DCGM_FT_INT64_BLANK + 3)
+	DCGM_FT_FP64_BLANK             = 140737488355328.0
+	DCGM_FT_FP64_NOT_FOUND         = float64(DCGM_FT_FP64_BLANK + 1.0)
+	DCGM_FT_FP64_NOT_SUPPORTED     = float64(DCGM_FT_FP64_BLANK + 2.0)
+	DCGM_FT_FP64_NOT_PERMISSIONED  = float64(DCGM_FT_FP64_BLANK + 3.0)
+	DCGM_FT_STR_BLANK              = "<<<NULL>>>"
+	DCGM_FT_STR_NOT_FOUND          = "<<<NOT_FOUND>>>"
+	DCGM_FT_STR_NOT_SUPPORTED      = "<<<NOT_SUPPORTED>>>"
+	DCGM_FT_STR_NOT_PERMISSIONED   = "<<<NOT_PERM>>>"
 
 	DCGM_FI_UNKNOWN                               = 0
 	DCGM_FI_DRIVER_VERSION                        = 1

pkg/gpu_collector.go

@@ -74,7 +74,7 @@ func ToMetric(values []dcgm.FieldValue_v1, c []Counter, d dcgm.Device) []Metric
 	for i, val := range values {
 		v := ToString(val)
 		// Filter out counters with no value
-		if v == "9223372036854775794" {
+		if v == SkipDCGMValue {
 			continue
 		}
 		m := Metric{
@@ -94,6 +94,34 @@ func ToMetric(values []dcgm.FieldValue_v1, c []Counter, d dcgm.Device) []Metric
 }
 
 func ToString(value dcgm.FieldValue_v1) string {
+	switch v := value.Int64(); v {
+	case dcgm.DCGM_FT_INT32_BLANK:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_INT32_NOT_FOUND:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_INT32_NOT_SUPPORTED:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_INT32_NOT_PERMISSIONED:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_INT64_BLANK:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_INT64_NOT_FOUND:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_INT64_NOT_SUPPORTED:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_INT64_NOT_PERMISSIONED:
+		return SkipDCGMValue
+	}
+	switch v := value.Float64(); v {
+	case dcgm.DCGM_FT_FP64_BLANK:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_FP64_NOT_FOUND:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_FP64_NOT_SUPPORTED:
+		return SkipDCGMValue
+	case dcgm.DCGM_FT_FP64_NOT_PERMISSIONED:
+		return SkipDCGMValue
+	}
 	switch v := value.FieldType; v {
 	case dcgm.DCGM_FT_STRING:
 		return value.String()

pkg/types.go

@@ -25,6 +25,7 @@ import (
 )
 
 var (
+	SkipDCGMValue   = "SKIPPING DCGM VALUE"
 	FailedToConvert = "ERROR - FAILED TO CONVERT TO STRING"
 
 	nvidiaResourceName = "nvidia.com/gpu"