feat: enable limiting and reserving gpu resources

plugins/resource/resource.go

@@ -15,6 +15,7 @@ type Resource struct {
 	Network        string `json:"network"`
 	NetworkIngress string `json:"network-ingress"`
 	NetworkEgress  string `json:"network-egress"`
+	NvidiaGPU      string `json:"nvidia-gpu"`
 }
 
 // ReportSingleApp is an internal function that displays the app report for one or more apps

plugins/resource/src/subcommands/subcommands.go

@@ -26,6 +26,7 @@ func main() {
 		network := args.String("network", "", "network: The amount of network bandwidth to limit processes by")
 		networkIngress := args.String("network-ingress", "", "network-ingress: The amount of ingress network bandwidth to limit processes by")
 		networkEgress := args.String("network-egress", "", "network-egress: The amount of egress network bandwidth to limit processes by")
+		nvidiaGpu := args.String("nvidia-gpu", "", "nvidia-gpu: The number of Nvidia GPUs to limit a process to")
 		args.Parse(os.Args[2:])
 
 		resources := resource.Resource{
@@ -35,6 +36,7 @@ func main() {
 			Network:        *network,
 			NetworkIngress: *networkIngress,
 			NetworkEgress:  *networkEgress,
+			NvidiaGPU:      *nvidiaGpu,
 		}
 
 		err = resource.CommandLimit(args.Args(), *processType, resources)
@@ -57,6 +59,7 @@ func main() {
 		network := args.String("network", "", "network: The amount of network bandwidth to reserve for processes")
 		networkIngress := args.String("network-ingress", "", "network-ingress: The amount of ingress network bandwidth to reserve for processes")
 		networkEgress := args.String("network-egress", "", "network-egress: The amount of egress network bandwidth to reserve for processes")
+		nvidiaGpu := args.String("nvidia-gpu", "", "nvidia-gpu: The number of Nvidia GPUs to resource for a process")
 		args.Parse(os.Args[2:])
 
 		resources := resource.Resource{
@@ -66,6 +69,7 @@ func main() {
 			Network:        *network,
 			NetworkIngress: *networkIngress,
 			NetworkEgress:  *networkEgress,
+			NvidiaGPU:      *nvidiaGpu,
 		}
 
 		err = resource.CommandReserve(args.Args(), *processType, resources)

plugins/resource/subcommands.go

@@ -100,6 +100,7 @@ func clearByResourceType(appName string, processType string, resourceType string
 			"network",
 			"network-ingress",
 			"network-egress",
+			"nvidia-gpu",
 		}
 
 		for _, key := range resources {
@@ -121,6 +122,7 @@ func setResourceType(appName string, processType string, r Resource, resourceTyp
 		"network":         r.Network,
 		"network-ingress": r.NetworkIngress,
 		"network-egress":  r.NetworkEgress,
+		"nvidia-gpu":      r.NvidiaGPU,
 	}
 
 	hasValues := false
@@ -182,6 +184,7 @@ func reportResourceType(appName string, processType string, resourceType string)
 		"network",
 		"network-ingress",
 		"network-egress",
+		"nvidia-gpu",
 	}
 
 	for _, key := range resources {

plugins/resource/triggers.go

@@ -32,6 +32,7 @@ func TriggerDockerArgsProcessDeploy(appName string, processType string) error {
 
 	validLimits := map[string]bool{
 		"cpu":         true,
+		"nvidia-gpu":  true,
 		"memory":      true,
 		"memory-swap": true,
 	}
@@ -54,6 +55,10 @@ func TriggerDockerArgsProcessDeploy(appName string, processType string) error {
 					parts[1] = "cpus"
 				}
 
+				if parts[1] == "nvidia-gpu" {
+					parts[1] = "gpus"
+				}
+
 				limits[parts[1]] = value
 			}
 			if parts[0] == "reserve" {