k8spod使用gpu

k8s pod使用gpu前提

1. k8s节点有gpu显卡
2. k8s节点安装gpu显卡驱动
3. k8s节点docker或containerd运行时使用nvidia-container-runtime
4. k8s部署gpu device plugin daemonset

1.安装gpu显卡驱动

查看节点显卡类型

nvidia-smi  -L
GPU 0: Tesla V100-SXM2-32GB (UUID: GPU-f2b15a66-0630-5f77-1f17-28abb3854f1c)

# 忘记没安装驱动，用不了上面命令，使用
lspci | grep -i nvidia
00:03.0 3D controller: NVIDIA Corporation Device 1eb8 (rev a1)
00:04.0 3D controller: NVIDIA Corporation Device 1eb8 (rev a1) 

根据 1eb8到这个网站查
http://pci-ids.ucw.cz/mods/PC/10de?action=help?help=pci

根据型号到这个网站查找驱动安装程序
https://www.nvidia.com/Download/Find.aspx#

# 下载
wget https://us.download.nvidia.com/tesla/515.65.01/NVIDIA-Linux-x86_64-515.65.01.run

chmod +x NVIDIA-Linux-x86_64-515.65.01.run

# 上述安装程序依赖这些包，安装
apt install gcc linux-kernel-headers dkms
sh NVIDIA-Linux-x86_64-515.65.01.run --ui=none --disable-nouveau --no-install-libglvnd --dkms -s

# 使用下面命令验证是否安装成功。
nvidia-smi
Thu Nov  3 19:17:50 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 515.65.01    Driver Version: 515.65.01    CUDA Version: 11.7     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla V100-SXM2...  On   | 00000000:00:08.0 Off |                    0 |
| N/A   36C    P0    37W / 300W |      4MiB / 32768MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      1499      G   /usr/lib/xorg/Xorg                  4MiB |
+-----------------------------------------------------------------------------+

2. 安装nvidia-container-runtime

curl -s -L https://nvidia.github.io/nvidia-container-runtime/gpgkey | sudo apt-key add -

curl -s -L https://nvidia.github.io/nvidia-container-runtime/$(. /etc/os-release;echo $ID$VERSION_ID)/nvidia-container-runtime.list | sudo tee /etc/apt/sources.list.d/nvidia-container-runtime.list

apt update

apt install nvidia-container-runtime -y

2.1修改默认运行时

2.2 cri为docker

修改/etc/docker/daemon.json，增加default-runtime,runtimes配置.

{
  "default-runtime": "nvidia",
  "runtimes": {
      "nvidia": {
          "path": "/usr/bin/nvidia-container-runtime",
          "runtimeArgs": []
      }
  },
  "registry-mirrors": [
     "https://docker.mirrors.ustc.edu.cn/",
     "https://hub-mirror.c.163.com/"
  ],
  "max-concurrent-downloads": 10,
  "log-driver": "json-file",
  "log-level": "warn",
  "log-opts": {
    "max-size": "10m",
    "max-file": "3"
    },
  "insecure-registries":
        ["127.0.0.1","192.168.12.12:8888"],
  "data-root":"/data/docker",
  "features":{"buildkit": true}
}

重启
systemctl restart docker

2.2 cri为containerd

修改/etc/containerd/config.toml，如果文件不存在
生成默认配置文件

mkdir /etc/containerd
containerd config default > /etc/containerd/config.toml
vi /etc/containerd/config.toml
...
    [plugins."io.containerd.grpc.v1.cri".containerd]
      snapshotter = "overlayfs"
      default_runtime_name = "runc"
      no_pivot = false
...
      [plugins."io.containerd.grpc.v1.cri".containerd.runtimes]
        [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]
          runtime_type = "io.containerd.runtime.v1.linux" # 将此处 runtime_type 的值改成 io.containerd.runtime.v1.linux
...
  [plugins."io.containerd.runtime.v1.linux"]
    shim = "containerd-shim"
    runtime = "nvidia-container-runtime" # 将此处 runtime 的值改成 nvidia-container-runtime
...

重启
systemctl restart containerd

3. 部署nivdia-device-plugin

kubectl apply -f nvidia-device-plugin.yaml

# Copyright (c) 2019, NVIDIA CORPORATION.  All rights reserved.
#
# 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.

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: nvidia-device-plugin-daemonset
  namespace: kube-system
spec:
  selector:
    matchLabels:
      name: nvidia-device-plugin-ds
  updateStrategy:
    type: RollingUpdate
  template:
    metadata:
      # This annotation is deprecated. Kept here for backward compatibility
      # See https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/
      annotations:
        scheduler.alpha.kubernetes.io/critical-pod: ""
      labels:
        name: nvidia-device-plugin-ds
    spec:
      tolerations:
      # This toleration is deprecated. Kept here for backward compatibility
      # See https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/
      - key: CriticalAddonsOnly
        operator: Exists
      - key: nvidia.com/gpu
        operator: Exists
        effect: NoSchedule
      # Mark this pod as a critical add-on; when enabled, the critical add-on
      # scheduler reserves resources for critical add-on pods so that they can
      # be rescheduled after a failure.
      # See https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/
      priorityClassName: "system-node-critical"
      containers:
      - image: nvidia/k8s-device-plugin:v0.7.1
        name: nvidia-device-plugin-ctr
        args: ["--fail-on-init-error=false"]
        securityContext:
          allowPrivilegeEscalation: false
          capabilities:
            drop: ["ALL"]
        volumeMounts:
          - name: device-plugin
            mountPath: /var/lib/kubelet/device-plugins
      volumes:
        - name: device-plugin
          hostPath:
            path: /var/lib/kubelet/device-plugins

4. 查看kubelet是否识别gpu

查看pod是否正常启动
kubectl get pod -n kube-system -o wide

describe node查看是否识别gpu
kubectl describe node vm-1-5-ubuntu

测试gpu pod启动
kubectl apply -f gpu-pod.yaml

apiVersion: v1
kind: Pod
metadata:
  name: gpu-pod
spec:
  containers:
    - name: cuda-container
      image: nvidia/cuda:9.0-devel
      resources:
        limits:
          nvidia.com/gpu: 1
    - name: digits-container
      image: nvidia/digits:6.0
      resources:
        limits:
          nvidia.com/gpu: 1

5. gpu共享

以上方式为独占gpu，GPU资源在节点上是以个数暴露给kubernetes集群来进行调度的，也就是说如果有两个后端应用pod需要使用到GPU资源，但集群节点上只有一张GPU物理卡的情况下，会导致两个后端应用容器中仅有一个可以正常运行，另一个pod则会处于pending状态。

gpu共享的配置阿里云和ucloud等都有对应文档配置，非云集群网上也有开源解决方案。