feat: support service.gpus and reservations.devices (#156)

* feat: support compose.yaml `service.gpus` and `service.deploy.resources.reservations.devices` DeviceRequests to Docker container
* Fix lint error
* Adjust test comments to work with linter
* Rename DeviceRequests to DeviceReservations to be more consistent with compose nomenclature
* Device reservation changes are immutable and should trigger a container recreate rather than update
This commit is contained in:
Justin Bradford
2025-10-27 16:49:38 +10:00
committed by GitHub
parent 0153e9f9e2
commit 3b4bcb7932
7 changed files with 653 additions and 1 deletions
+6
View File
@@ -83,11 +83,17 @@ func EvalContainerSpecChange(current api.ServiceSpec, new api.ServiceSpec) Conta
}
}
// Device reservations are immutable, so we'll need to recreate if any have changed
if !reflect.DeepEqual(current.Container.Resources.DeviceReservations, newResources.DeviceReservations) {
return ContainerNeedsRecreate
}
// Check if any mutable properties changed.
if !current.Caddy.Equals(new.Caddy) {
return ContainerNeedsRecreate
}
// Remaining resources are mutable.
if !reflect.DeepEqual(current.Container.Resources, newResources) {
return ContainerNeedsUpdate
}
+419
View File
@@ -3,6 +3,7 @@ package deploy
import (
"testing"
"github.com/docker/docker/api/types/container"
"github.com/docker/docker/api/types/mount"
"github.com/psviderski/uncloud/pkg/api"
"github.com/stretchr/testify/assert"
@@ -1350,6 +1351,398 @@ func TestEvalContainerSpecChange_Volumes(t *testing.T) {
}
}
func TestEvalContainerSpecChange_DeviceReservations(t *testing.T) {
t.Parallel()
tests := []struct {
name string
current api.ContainerResources
new api.ContainerResources
want ContainerSpecStatus
}{
{
name: "empty",
current: api.ContainerResources{},
new: api.ContainerResources{},
want: ContainerUpToDate,
},
{
name: "both nil",
current: api.ContainerResources{DeviceReservations: nil},
new: api.ContainerResources{DeviceReservations: nil},
want: ContainerUpToDate,
},
{
name: "both empty",
current: api.ContainerResources{DeviceReservations: []container.DeviceRequest{}},
new: api.ContainerResources{DeviceReservations: []container.DeviceRequest{}},
want: ContainerUpToDate,
},
{
name: "set GPU request",
current: api.ContainerResources{},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: -1, // all
Capabilities: [][]string{{"gpu"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "unset GPU request",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: -1,
Capabilities: [][]string{{"gpu"}},
},
},
},
new: api.ContainerResources{},
want: ContainerNeedsRecreate,
},
{
name: "identical GPU request",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "nvidia",
Count: 2,
Capabilities: [][]string{{"gpu", "compute"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "nvidia",
Count: 2,
Capabilities: [][]string{{"gpu", "compute"}},
},
},
},
want: ContainerUpToDate,
},
{
name: "change GPU count",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 2,
Capabilities: [][]string{{"gpu"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "change GPU driver",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "nvidia",
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "amd",
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "change device IDs",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
DeviceIDs: []string{"0"},
Capabilities: [][]string{{"gpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
DeviceIDs: []string{"0", "1"},
Capabilities: [][]string{{"gpu"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "change capabilities",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu", "compute"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "set options",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
Options: map[string]string{
"key": "value",
},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "change options",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
Options: map[string]string{
"key": "value1",
},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
Options: map[string]string{
"key": "value2",
},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "unset options",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
Options: map[string]string{
"key": "value",
},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "add device request",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
{
Count: 1,
Capabilities: [][]string{{"tpu"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "remove device request",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
{
Count: 1,
Capabilities: [][]string{{"tpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "multiple identical device requests",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "nvidia",
Count: 2,
Capabilities: [][]string{{"gpu", "compute"}},
},
{
Count: 1,
Capabilities: [][]string{{"tpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "nvidia",
Count: 2,
Capabilities: [][]string{{"gpu", "compute"}},
},
{
Count: 1,
Capabilities: [][]string{{"tpu"}},
},
},
},
want: ContainerUpToDate,
},
{
name: "reordered device requests",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "nvidia",
Count: 2,
Capabilities: [][]string{{"gpu"}},
},
{
Count: 1,
Capabilities: [][]string{{"tpu"}},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"tpu"}},
},
{
Driver: "nvidia",
Count: 2,
Capabilities: [][]string{{"gpu"}},
},
},
},
want: ContainerNeedsRecreate,
},
{
name: "complex GPU configuration identical",
current: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "nvidia",
Count: 2,
DeviceIDs: []string{"0", "1"},
Capabilities: [][]string{{"gpu", "compute", "utility"}},
Options: map[string]string{
"runtime": "nvidia",
"compute": "exclusive",
},
},
},
},
new: api.ContainerResources{
DeviceReservations: []container.DeviceRequest{
{
Driver: "nvidia",
Count: 2,
DeviceIDs: []string{"0", "1"},
Capabilities: [][]string{{"gpu", "compute", "utility"}},
Options: map[string]string{
"runtime": "nvidia",
"compute": "exclusive",
},
},
},
},
want: ContainerUpToDate,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
currentSpec := api.ServiceSpec{
Container: api.ContainerSpec{
Image: "nvidia/cuda:latest",
Resources: tt.current,
},
}
newSpec := api.ServiceSpec{
Container: api.ContainerSpec{
Image: "nvidia/cuda:latest",
Resources: tt.new,
},
}
result := EvalContainerSpecChange(currentSpec, newSpec)
assert.Equal(t, tt.want, result)
})
}
}
func TestEvalContainerSpecChange_Mixed(t *testing.T) {
t.Parallel()
@@ -1399,6 +1792,32 @@ func TestEvalContainerSpecChange_Mixed(t *testing.T) {
},
want: ContainerNeedsRecreate,
},
{
name: "mutable memory change with immutable device reservation change",
current: api.ServiceSpec{
Container: api.ContainerSpec{
Image: "nvidia/cuda:latest",
Resources: api.ContainerResources{
Memory: 100 * 1024 * 1024,
},
},
},
new: api.ServiceSpec{
Container: api.ContainerSpec{
Image: "nvidia/cuda:latest",
Resources: api.ContainerResources{
Memory: 200 * 1024 * 1024,
DeviceReservations: []container.DeviceRequest{
{
Count: 1,
Capabilities: [][]string{{"gpu"}},
},
},
},
},
},
want: ContainerNeedsRecreate,
},
}
for _, tt := range tests {