Merge pull request kubernetes#42 from chentao1596/master

tasks-manage-gpus-scheduling-gpus-pr-2017-7-15

docs/tasks/manage-gpus/scheduling-gpus.md

@@ -1,34 +1,62 @@
 ---
 assignees:
 - vishh
+title: 调度 GPU
+redirect_from:
+- "/docs/user-guide/gpus/"
+- "/docs/user-guide/gpus.html"
+---
+<!--
+---
+assignees:
+- vishh
 title: Schedule GPUs
 redirect_from:
 - "/docs/user-guide/gpus/"
 - "/docs/user-guide/gpus.html"
 ---
+-->
 
 {% capture overview %}
 
+<!--
 Kubernetes includes **experimental** support for managing NVIDIA GPUs spread across nodes.
 This page describes how users can consume GPUs and the current limitations.
+-->
+Kubernetes 提供对分布在节点上的 NVIDIA GPU 进行管理的**实验**支持。本页描述用户如何使用 GPU 以及当前使用的一些限制
 
 {% endcapture %}
 
 {% capture prerequisites %}
 
+<!--
 1. Kubernetes nodes have to be pre-installed with Nvidia drivers. Kubelet will not detect Nvidia GPUs otherwise. Try to re-install nvidia drivers if kubelet fails to expose Nvidia GPUs as part of Node Capacity.
+-->
+1. Kubernetes 节点必须预先安装好 NVIDIA 驱动，否则，Kubelet 将检测不到可用的GPU信息；如果节点的 Capacity 属性中没有出现 NIVIDA GPU 的数量，有可能是驱动没有安装或者安装失败，请尝试重新安装
+<!--
 2. A special **alpha** feature gate `Accelerators` has to be set to true across the system: `--feature-gates="Accelerators=true"`.
+-->
+2. 在整个 Kubernetes 系统中，feature-gates 里面特定的 **alpha** 特性参数 `Accelerators` 必须设置为 true：`--feature-gates="Accelerators=true"`
+<!--
 3. Nodes must be using `docker engine` as the container runtime.
+-->
+3. Kuberntes 节点必须使用 `docker` 引擎作为容器的运行引擎
 
+<!--
 The nodes will automatically discover and expose all Nvidia GPUs as a schedulable resource.
+-->
+上述预备工作完成后，节点会自动发现它上面的 NVIDIA GPU，并将其作为可调度资源暴露
 
 {% endcapture %}
 
 {% capture steps %}
 
 ## API
 
+<!--
 Nvidia GPUs can be consumed via container level resource requirements using the resource name `alpha.kubernetes.io/nvidia-gpu`.
+-->
+容器可以通过名称为 `alpha.kubernetes.io/nvidia-gpu` 的标识来申请需要使用的 NVIDIA GPU 的数量
 
 ```yaml
 apiVersion: v1
@@ -47,16 +75,37 @@ spec:
           alpha.kubernetes.io/nvidia-gpu: 3 # requesting 3 GPUs
 ```
 
+<!--
 - GPUs can be specified in the `limits` section only.
+-->
+- GPU 只能在容器资源的 `limits` 中配置
+<!--
 - Containers (and pods) do not share GPUs.
+-->
+- 容器和 Pod 都不支持共享 GPU
+<!--
 - Each container can request one or more GPUs.
+-->
+- 每个容器可以申请使用一个或者多个 GPU
+<!--
 - It is not possible to request a portion of a GPU.
+-->
+- GPU 必须以整数为单位被申请使用
+<!--
 - Nodes are expected to be homogenous, i.e. run the same GPU hardware.
+-->
+- 所有节点的 GPU 硬件要求相同
 
+<!--
 If your nodes are running different versions of GPUs, then use Node Labels and Node Selectors to schedule pods to appropriate GPUs.
 Following is an illustration of this workflow:
+-->
+如果在不同的节点上面安装了不同版本的 GPU，可以通过设置节点标签以及使用节点选择器的方式将 pod 调度到期望运行的节点上。工作流程如下：
 
+<!--
 As part of your Node bootstrapping, identify the GPU hardware type on your nodes and expose it as a node label.
+-->
+在节点上，识别出 GPU 硬件类型，然后将其作为节点标签进行暴露
 
 ```shell
 NVIDIA_GPU_NAME=$(nvidia-smi --query-gpu=gpu_name --format=csv,noheader --id=0)
@@ -65,7 +114,10 @@ KUBELET_OPTS="$KUBELET_OPTS --node-labels='alpha.kubernetes.io/nvidia-gpu-name=$
 echo "KUBELET_OPTS=$KUBELET_OPTS" > /etc/default/kubelet
 ```
 
+<!--
 Specify the GPU types a pod can use via [Node Affinity](/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity) rules.
+-->
+在 pod 上，通过节点[亲和性](/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity)规则为它指定可以使用的 GPU 类型
 
 ```yaml
 kind: pod
@@ -99,19 +151,40 @@ spec:
           alpha.kubernetes.io/nvidia-gpu: 2
 ```
 
+<!--
 This will ensure that the pod will be scheduled to a node that has a `Tesla K80` or a `Tesla P100` Nvidia GPU.
+-->
+上述设定可以确保 pod 会被调度到包含名称为 `alpha.kubernetes.io/nvidia-gpu-name` 的标签并且标签的值为 `Tesla K80` 或者 `Tesla P100` 的节点上
 
+<!--
 ### Warning
+-->
+### 警告
 
+<!--
 The API presented here **will change** in an upcoming release to better support GPUs, and hardware accelerators in general, in Kubernetes.
+-->
+当未来的 Kubernetes 版本能够更好的支持GPU以及一般的硬件加速器时，这里的 API 描述**将会随之做出变更**
 
+<!--
 ## Access to CUDA libraries
+-->
+## 访问 CUDA 库
 
+<!--
 As of now, CUDA libraries are expected to be pre-installed on the nodes.
+-->
+到目前为止，还需要预先在节点上安装 CUDA 库
 
+<!--
 To mitigate this, you can copy the libraries to a more permissive folder in ``/var/lib/`` or change the permissions directly. (Future releases will automatically perform this operation)
+-->
+为了避免后面使用库出现问题，可以将库放到 ``/var/lib/`` 下的某个文件夹下，或者直接改变库目录的权限(以后的版本会自动完成这一过程)
 
+<!--
 Pods can access the libraries using `hostPath` volumes.
+-->
+Pods能够通过 `hostPath` 卷来访问库
 
 ```yaml
 kind: Pod
@@ -140,13 +213,31 @@ spec:
     name: lib
 ```
 
+<!--
 ## Future
+-->
+## 未来
 
+<!--
 - Support for hardware accelerators is in it's early stages in Kubernetes.
+-->
+- Kubernetes 对硬件加速器的支持还处在早期阶段
+<!--
 - GPUs and other accelerators will soon be a native compute resource across the system.
+-->
+- GPU 和其它的加速器很快会成为系统的本地计算资源
+<!--
 - Better APIs will be introduced to provision and consume accelerators in a scalable manner.
+-->
+- 将引入更好的 API 以可扩展的方式提供和使用加速器
+<!--
 - Kubernetes will automatically ensure that applications consuming GPUs gets the best possible performance.
+-->
+- Kubernets 将会自动确保应用在使用 GPU 时得到最佳性能
+<!--
 - Key usability problems like access to CUDA libraries will be addressed.
+-->
+- 类似访问 CUDA 库这种关键的可用性问题将得到解决
 
 {% endcapture %}