diff --git a/plugins/device/cmd/nvidia/device.go b/plugins/device/cmd/nvidia/device.go
index 84b22e5da..2613d8e77 100644
--- a/plugins/device/cmd/nvidia/device.go
+++ b/plugins/device/cmd/nvidia/device.go
@@ -3,6 +3,7 @@ package nvidia
 import (
 	"context"
 	"fmt"
+	"strings"
 	"sync"
 	"time"
 
@@ -29,6 +30,11 @@ const (
 	notAvailable = "N/A"
 )
 
+const (
+	// Nvidia-container-runtime environment variable names
+	nvidiaVisibleDevices = "NVIDIA_VISIBLE_DEVICES"
+)
+
 var (
 	// pluginInfo describes the plugin
 	pluginInfo = &base.PluginInfoResponse{
@@ -149,9 +155,50 @@ func (d *NvidiaDevice) Fingerprint(ctx context.Context) (<-chan *device.Fingerpr
 	return outCh, nil
 }
 
-// Reserve returns information on how to mount the given devices.
+type reservationError struct {
+	notExistingIDs []string
+}
+
+func (e *reservationError) Error() string {
+	return fmt.Sprintf("unknown device IDs: %s", strings.Join(e.notExistingIDs, ","))
+}
+
+// Reserve returns information on how to mount given devices.
+// Assumption is made that nomad server is responsible for correctness of
+// GPU allocations, handling tricky cases such as double-allocation of single GPU
 func (d *NvidiaDevice) Reserve(deviceIDs []string) (*device.ContainerReservation, error) {
-	return nil, nil
+	if len(deviceIDs) == 0 {
+		return &device.ContainerReservation{}, nil
+	}
+	// Due to the asynchronous nature of NvidiaPlugin, there is a possibility
+	// of race condition
+	//
+	// Timeline:
+	// 	1 - fingerprint reports that GPU with id "1" is present
+	//  2 - the following events happen at the same time:
+	// 		a) server decides to allocate GPU with id "1"
+	//      b) fingerprint check reports that GPU with id "1" is no more present
+	//
+	// The latest and always valid version of fingerprinted ids are stored in
+	// d.devices map. To avoid this race condition an error is returned if
+	// any of provided deviceIDs is not found in d.devices map
+	d.deviceLock.RLock()
+	var notExistingIDs []string
+	for _, id := range deviceIDs {
+		if _, deviceIDExists := d.devices[id]; !deviceIDExists {
+			notExistingIDs = append(notExistingIDs, id)
+		}
+	}
+	d.deviceLock.RUnlock()
+	if len(notExistingIDs) != 0 {
+		return nil, &reservationError{notExistingIDs}
+	}
+
+	return &device.ContainerReservation{
+		Envs: map[string]string{
+			nvidiaVisibleDevices: strings.Join(deviceIDs, ","),
+		},
+	}, nil
 }
 
 // Stats streams statistics for the detected devices.
diff --git a/plugins/device/cmd/nvidia/device_test.go b/plugins/device/cmd/nvidia/device_test.go
index 43ef1f570..b1fa4b17a 100644
--- a/plugins/device/cmd/nvidia/device_test.go
+++ b/plugins/device/cmd/nvidia/device_test.go
@@ -1,7 +1,13 @@
 package nvidia
 
 import (
+	"testing"
+
 	"github.com/hashicorp/nomad/plugins/device/cmd/nvidia/nvml"
+
+	hclog "github.com/hashicorp/go-hclog"
+	"github.com/hashicorp/nomad/plugins/device"
+	"github.com/stretchr/testify/require"
 )
 
 type MockNvmlClient struct {
@@ -19,3 +25,91 @@ func (c *MockNvmlClient) GetFingerprintData() (*nvml.FingerprintData, error) {
 func (c *MockNvmlClient) GetStatsData() ([]*nvml.StatsData, error) {
 	return c.StatsResponseReturned, c.StatsError
 }
+
+func TestReserve(t *testing.T) {
+	for _, testCase := range []struct {
+		Name                string
+		ExpectedReservation *device.ContainerReservation
+		ExpectedError       error
+		Device              *NvidiaDevice
+		RequestedIDs        []string
+	}{
+		{
+			Name:                "All RequestedIDs are not managed by Device",
+			ExpectedReservation: nil,
+			ExpectedError: &reservationError{[]string{
+				"UUID1",
+				"UUID2",
+				"UUID3",
+			}},
+			RequestedIDs: []string{
+				"UUID1",
+				"UUID2",
+				"UUID3",
+			},
+			Device: &NvidiaDevice{
+				logger: hclog.NewNullLogger(),
+			},
+		},
+		{
+			Name:                "Some RequestedIDs are not managed by Device",
+			ExpectedReservation: nil,
+			ExpectedError: &reservationError{[]string{
+				"UUID1",
+				"UUID2",
+			}},
+			RequestedIDs: []string{
+				"UUID1",
+				"UUID2",
+				"UUID3",
+			},
+			Device: &NvidiaDevice{
+				devices: map[string]struct{}{
+					"UUID3": {},
+				},
+				logger: hclog.NewNullLogger(),
+			},
+		},
+		{
+			Name: "All RequestedIDs are managed by Device",
+			ExpectedReservation: &device.ContainerReservation{
+				Envs: map[string]string{
+					nvidiaVisibleDevices: "UUID1,UUID2,UUID3",
+				},
+			},
+			ExpectedError: nil,
+			RequestedIDs: []string{
+				"UUID1",
+				"UUID2",
+				"UUID3",
+			},
+			Device: &NvidiaDevice{
+				devices: map[string]struct{}{
+					"UUID1": {},
+					"UUID2": {},
+					"UUID3": {},
+				},
+				logger: hclog.NewNullLogger(),
+			},
+		},
+		{
+			Name:                "No IDs requested",
+			ExpectedReservation: &device.ContainerReservation{},
+			ExpectedError:       nil,
+			RequestedIDs:        nil,
+			Device: &NvidiaDevice{
+				devices: map[string]struct{}{
+					"UUID1": {},
+					"UUID2": {},
+					"UUID3": {},
+				},
+				logger: hclog.NewNullLogger(),
+			},
+		},
+	} {
+		actualReservation, actualError := testCase.Device.Reserve(testCase.RequestedIDs)
+		req := require.New(t)
+		req.Equal(testCase.ExpectedReservation, actualReservation)
+		req.Equal(testCase.ExpectedError, actualError)
+	}
+}