kubernetes/test/e2e/framework/gpu_util.go
Kubernetes Submit Queue 459b8f65eb
Merge pull request #59230 from tianshapjq/parse-dp-image
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.

should log error when error in parsing device plugin image

**What this PR does / why we need it**:
add some extra error log in parsing the device plugin image. Error happened in my setup when fetching the yaml from url, but since no error log printed it cost a long time to figure out the reason. So it'd be nice to print the error IMO.

**Special notes for your reviewer**:

**Release note**:

```release-note
NONE
```

/sig-node
2018-08-22 22:56:09 -07:00

87 lines
2.6 KiB
Go

/*
Copyright 2017 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package framework
import (
"k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/util/uuid"
"github.com/golang/glog"
. "github.com/onsi/gomega"
)
const (
// GPUResourceName is the extended name of the GPU resource since v1.8
// this uses the device plugin mechanism
NVIDIAGPUResourceName = "nvidia.com/gpu"
// TODO: Parametrize it by making it a feature in TestFramework.
// so we can override the daemonset in other setups (non COS).
// GPUDevicePluginDSYAML is the official Google Device Plugin Daemonset NVIDIA GPU manifest for GKE
GPUDevicePluginDSYAML = "https://raw.githubusercontent.com/kubernetes/kubernetes/master/cluster/addons/device-plugins/nvidia-gpu/daemonset.yaml"
)
// TODO make this generic and not linked to COS only
// NumberOfGPUs returs the number of GPUs advertised by a node
// This is based on the Device Plugin system and expected to run on a COS based node
// After the NVIDIA drivers were installed
func NumberOfNVIDIAGPUs(node *v1.Node) int64 {
val, ok := node.Status.Capacity[NVIDIAGPUResourceName]
if !ok {
return 0
}
return val.Value()
}
// NVIDIADevicePlugin returns the official Google Device Plugin pod for NVIDIA GPU in GKE
func NVIDIADevicePlugin() *v1.Pod {
ds, err := DsFromManifest(GPUDevicePluginDSYAML)
Expect(err).NotTo(HaveOccurred())
p := &v1.Pod{
ObjectMeta: metav1.ObjectMeta{
Name: "device-plugin-nvidia-gpu-" + string(uuid.NewUUID()),
Namespace: metav1.NamespaceSystem,
},
Spec: ds.Spec.Template.Spec,
}
// Remove node affinity
p.Spec.Affinity = nil
return p
}
func GetGPUDevicePluginImage() string {
ds, err := DsFromManifest(GPUDevicePluginDSYAML)
if err != nil {
glog.Errorf("Failed to parse the device plugin image: %v", err)
return ""
}
if ds == nil {
glog.Errorf("Failed to parse the device plugin image: the extracted DaemonSet is nil")
return ""
}
if len(ds.Spec.Template.Spec.Containers) < 1 {
glog.Errorf("Failed to parse the device plugin image: cannot extract the container from YAML")
return ""
}
return ds.Spec.Template.Spec.Containers[0].Image
}