pso-csi:PSO CSI掌舵图

  • w9_392808
    了解作者
  • 1MB
    文件大小
  • zip
    文件格式
  • 0
    收藏次数
  • VIP专享
    资源类型
  • 0
    下载次数
  • 2022-05-16 11:30
    上传日期
纯服务协调器(PSO)CSI驱动程序 使用的是Google Anthos还是OpenShift 3.11? 请改用 什么是PSO? Pure Service Orchestrator(PSO)为容器提供了存储即服务,从而为开发人员提供了公共云的敏捷性以及本地基础结构的可靠性和安全性。 智能配置PSO通过评估多个因素(例如性能负载,阵列的容量和运行状况以及策略标签),自动针对每个存储请求实时做出最佳配置决策。 弹性缩放PSO将所有Pure FlashArray:trade_mark:和FlashBlade:trade_mark:阵列整合在一个共享的基础架构上,并根据需要支持文件和块,PSO使得添加新阵列变得毫不费力,因此您可以随着环境的增长而扩展。 透明恢复为了确保您的服务保持稳定,PSO会自我修复–因此,您可以防止由于节点故障,阵列性能限制和磁盘空间不足等问题而导致的数据损坏。 功能列表 必备软件 请通读所有这些内容! 自
pso-csi-master.zip
内容介绍
# pure-pso This helm chart installs the Pure Service Orchestrator CSI plugin on a Kubernetes cluster. ## Important Notes 1. **Please create a new values.yaml file for PSO 6.x!** The format of the values file has changed since PSO 5.x and it is imperative you account for these differences. 2. Pure Service Orchestrator deploys a CockroachDB datastore replicated across the provided storage backends. More information on how the datastore works can be found [here](../docs/pso-datastore.md). 3. By default, PSO sets CockroachDB datastore to consume 50% of memory per node, which is recommended by CockroachDB. However, you can use [database.resources](https://github.com/purestorage/pso-csi/blob/master/pure-pso/values.yaml#L173-L176) in the `values.yaml` to constrain the CPU and memory usage. 4. Currently, there is **no upgrade supported** from previous versions that do not deploy the datastore (PSO 5.x and lower). 5. You **MUST** supply a unique `clusterID` in values.yaml. This was previously called `namespace.pure`. `clusterID` must be less than or equal to 22 characters in length. `clusterID` must be unique between **all** Kubernetes clusters using your Pure devices or naming conflicts will result. **WARNING** Do not change `clusterID` once it has been set during the initial installation of PSO on a cluster. 6. `helm uninstall` will perform the initial uninstallation, but some pods will continue to clean up post-installation. They should go away after cleanup is complete. 7. Note that PSO CSI only supports the Beta version snapshotter APIs. The snapshotter CRDs for the Beta version APIs have been upgraded, therefore use only release-2.0 CRDs as detailed below. In addition, the VolumeSnapshotClass pure-snapshotclass will be automatically applied to the cluster on Helm install if the cluster already has the snapshot Beta CRD defined. 8. Ensure IPv4 forwarding is enabled on all nodes to allow inter-node communication. See here for more [details](../docs/ipv4_forwarding.md) 9. **An NTP implementation (such as ntpd or chronyd) must be installed and running on all Kubernetes cluster nodes** 10. PSO 6.x requires at least 3+ nodes running the database, and 5+ nodes is recommended. They may run other workloads (they don't have to be dedicated), but for fault tolerance, the database will be spread across these nodes. 11. **[For Kubernetes version less than 1.17.6/1.18.6 please refer to this link/issue when using vxlan with Flannel or Calico](https://github.com/kubernetes/kubernetes/issues/87852).** You may experience numerous `CrashLoopBackoff` problems if you encounter this issue. ## Using controller attach-detach or restricting plugin pods to nodes More details on setting up controller attach-detach, or on restricting various plugin components to specific pods such as database nodes, can be found [here](../docs/csi-controller-attach-detach.md). ## CSI Snapshot and Clone features for Kubernetes More details on using the snapshot and clone functionality can be found [here](../docs/csi-snapshot-clones.md). ## Using Per-Volume FileSystem Options with Kubernetes More details on using customized filesystem options can be found [here](../docs/csi-filesystem-options.md). ## Using Read-Write-Many (RWX) volumes with Kubernetes More details on using Read-Write-Many (RWX) volumes with Kubernetes can be found [here](../docs/csi-read-write-many.md). ## PSO use of StorageClass There are some `StorageClass` definitions provided by the PSO installation. Refer [here](../docs/custom-storageclasses.md) for more details on these storage classes and how to create your own custom storage classes that can be used by PSO. ## Installation ### Upgrading from PSO 5.x Please use the instructions [here](../docs/csi-5to6-upgradepath.md). ### Configure NTP The PSO CSI driver requires all compute node clocks to be within 500ms. Ensure that an implementation of NTP is installaed and running on all cluster members, even those running as virtual machines. Example implementations include `ntp`, `chronyd`, `kvm-clock` and `system-timed` ### Install the plugin in a separate namespace (i.e. project) For security reasons, it is strongly recommended to install the plugin in a separate namespace/project. **Do not use the `default` namespace.** Make sure the namespace exists, otherwise create it before installing the plugin. ```bash kubectl create namespace <pso-namespace> ``` ### Configure Helm Add the Pure Storage PSO helm repository to your helm installation. ```bash helm repo add pure https://purestorage.github.io/pso-csi helm repo update helm search repo pure-pso -l ``` **Note: The chart name is case sensitive.** #### For offline installations (optional) Download the PSO helm chart ```bash git clone https://github.com/purestorage/pso-csi.git ``` Create and customize your own `values.yaml` and install the helm chart using this, and keep the file for future use. The easiest way is to copy the default [./values.yaml](https://raw.githubusercontent.com/purestorage/pso-csi/master/pure-pso/values.yaml) provided in the helm chart. ### OpenShift Pre-Requisites Please review the steps detailed [here](../docs/openshift_mc.md) for OpenShist specific additional steps. ### Dry run the installation This will validate your `values.yaml` and check it is working correctly. ```bash helm install pure-pso pure/pure-pso --version <version> --namespace <pso-namespace> -f <your_own_dir>/values.yaml --dry-run --debug ``` **Note: The `--version` flag is optional. Not providing this will install the latest GA version.** ### Run the Install ```bash helm install pure-pso pure/pure-pso --version <version> --namespace <pso-namespace> -f <your_own_dir>/values.yaml ``` **Note: The `--version` flag is optional. Not providing this will install the latest GA version.** The settings in your `values.yaml` overwrite the ones in `pure-pso/values.yaml` file, but any specified with the `--set` option applied to the install command will take precedence. For example ```bash helm install pure-pso pure/pure-pso --version <version> --namespace <pso-namespace> -f <your_own_dir>/values.yaml \ --set flasharray.sanType=fc \ --set clusterID=k8s_xxx ``` ### Post-Installation After installing, you should see pods like the following: ```bash > kubectl get pods -n <pso-namespace> NAME READY STATUS RESTARTS AGE pso-csi-controller-0 6/6 Running 0 52s pso-csi-node-bdr4m 3/3 Running 0 52s pso-csi-node-fr9c9 3/3 Running 0 52s pso-csi-node-sx6kp 3/3 Running 0 52s pso-db-0-0 1/1 Running 0 23s pso-db-1-0 1/1 Running 0 23s pso-db-2-0 1/1 Running 0 23s pso-db-3-0 1/1 Running 0 23s pso-db-4-0 1/1 Running 0 23s pso-db-cockroach-operator-5dbbc8855-sr2ks 1/1 Running 0 52s pso-db-deployer-56444bbb78-2tbsx 1/1 Running 0 52s ``` #### pso-csi-controller The CSI controller server is responsible for provisioning volumes, creating snapshots, and any other user-initiated action that requires a management API request to the storage arrays. Only one CSI controller should be running. #### pso-csi-node The CSI node server should be running on all compute nodes where volumes will be attached. The CSI node server does not make any management API requests to the storage arrays. #### pso-db The PSO database persists metadata for volumes and snapshots created by the PSO CSI driver. The data is replicated across the storage arrays for high availability. The `pso-db-deployer` and `pso-db-cockroach-operator` work in tandem to keep databas
评论
    相关推荐
    • SOA IBM Flash架构演示
      IBM flash 的小短片,经典讲述什么是SOA
    • 组织架构模板psd
      组织架构模板,psd多层格式,方便与制作ppt,企业架构
    • flash 殿堂之路
      flash 殿堂之路的pdf文件,不是扫描版本的
    • 精简flash播放器
      精简flash播放器,使用方便有有列表功能。播放界面可自定义
    • FLASH个人简历模板
      互动式个性简历,此为最基本架构,可添加音频,视频,FLASH等多元素的简历
    • mips nandflash驱动
      mips架构下nandflash驱动模板
    • freeBSD架构中文手册
      freeBSD官方的关于架构方面的用户手册 pdf版。200多页
    • Flex体系架构剖析
      Flex体系架构剖析!!感觉不错,推荐给大家!!
    • FLEX主页面架构
      首页显示自定义的功能窗口,窗口的位置相互可以拖拽调换,做项目的首页GOOD。如可以分为“代办”“已办”“消息”“邮件”等功能窗口,窗口拖来拖去赞一个哦。窗口还可以放大显示,代码比较多,研究吧。
    • spiflash 驱动代码
      完整的spiflash驱动代码,架构扩展性良好,兼容性广,支持四线串行闪存QPI模式。 先要调用init函数初始化,才能调用读写等函数。