CN1848787A - An automated rapid deployment method for cluster server system nodes - Google Patents

Abstract

A method for disposing nodes of cluster server system automatically and quickly includes obtaining image, defining object node, erecting operation system of object node, erecting and configuring application service of object node and configuring operation system identification of object node for realizing automation of erection course for cluster server system node.

Description

An automated rapid deployment method for cluster server system nodes

Technical field

The invention belongs to the technical field of computers, and relates to an automatic rapid deployment method for cluster server system nodes, in particular to a method for installing Unix and Linux operating systems and application software on the cluster server nodes. The present invention is also suitable for computer manufacturers and computer sales companies to use in the batch system pre-ordering process before products leave the factory or before sales.

Background technique

High-performance cluster computing systems are currently widely used, and the scale of clusters is also expanding. During the construction and operation of clusters, manual repeated installation and upgrade of operating systems and application software packages on cluster nodes will become a difficult task.

In the prior art prior to the present invention, the operating system and application software of cluster nodes are generally deployed by manual or ghost (hard disk duplication software developed by Symantec) hard disk recording or NFS shared network installation method. The manual installation method mainly uses the CD to install, and it takes a lot of time and manpower to install each node, and the node configuration process is very cumbersome and error-prone. In the engineering practice of building a cluster system, the ghost hard disk burning method and the NFS shared network installation method are commonly used: the ghost hard disk burning method needs to disassemble the hard disk of the installed node, and requires the target disk to be the same size as the source disk; NFS shared network installation The method needs to configure a node in the network as an NFS server, copy the files in the operating system installation disk to this node, and then make the installed node start through the network card or floppy disk, download the relevant files from the NFS server node and finish installation.

Some existing deployment software such as OSCAR (Open Source Cluster Application Resource, a set of cluster application software packages developed by Open Cluster Group) and SystemImager (system imager) have a small application range, single functions, and inconvenient operation.

Contents of Invention

In view of the above-mentioned technical status, the purpose of the present invention is to provide a new automatic rapid deployment method suitable for cluster server system nodes, which can simplify the installation work of the cluster server system and realize the automation of the installation process.

The technical solution for realizing the above-mentioned inventive method is: an automatic rapid deployment method for cluster server system nodes, characterized in that the method adopts a client/server architecture, which is configured with a deployment server node, a switch, a or multiple image server nodes, source nodes and target nodes of the software system to be installed;

The target node, the deployment server node and the image server node are all connected by Ethernet, using a star-shaped network topology model, and all target nodes must be directly connected to the deployment server node through a switch or hub; among them, the deployment server node is the The server side of the server is used to provide dynamic host configuration protocol service, simple file transfer protocol service, deployment service and configuration service;

When deploying the target node, first, start the target node from the network card, download the deployment agent from the deployment server node through the DHCP (Dynamic Host Configuration Protocol) protocol and TFTP (Trivial File Transfer Protocol) protocol, run the deployment agent in memory, and deploy After the agent communicates with the deployment server node, it completes the installation of the operating system on the target node; then, the deployment server node controls the target node to start from the local hard disk, runs the operating system just deployed, and the configuration agent completes the installation and operation of the application service image Configuration of system information; after configuration, the target node is a cluster node that can be used immediately.

The image server node is installed based on the disk image, and the operating system image and the application service image are deployed separately, among which:

The operating system image is an image file that includes the operating system that needs to be installed on the target node, and is generated by the deployment agent by obtaining a copy of the operating system of the source node through a disk image;

The application service image is an image file of the application service software that needs to be installed on the target node. It is uploaded by the administrator and generated by defining the relevant XML description file; both image files are stored on the image server node. According to the specific use purpose of the target node, the image is dynamically combined, and the target node accesses the image file on the image server node through the NFS file system.

When deploying server nodes to deploy operating system images to target nodes, use reliable multicast or unicast network transmission mode; use network file system NFS, preboot execution environment PXE (PrebootExecution Environment) protocol, dynamic host configuration protocol DHCP, simple file transfer The protocol TFTP and WOL (Wakeup On LAN) are used to realize the rapid and automatic deployment of the cluster system.

The advantage of the inventive method is:

A. Provide a single control point function. The administrator only needs to connect to the deployment server node through the terminal, and all control operations are completed by the deployment server node. The target node only needs to be set to start with the network card through the BIOS before deployment, and the deployment process is completely automated; the target node does not need peripheral devices such as keyboard, mouse, monitor, CD-ROM, and floppy drive.

B. Provide the acquisition and deployment functions of multi-type operating system images. Based on the installation method of the disk image, the disk operation tool is bound to the image acquisition agent, no matter what type of operating system is installed on the image source node, the operating system image can be obtained.

C. Using the deployment idea of separating the operating system image and the application service image, the administrator can dynamically combine images according to the specific purpose of use of the target node during the installation process of the cluster target node. The adaptability of the cluster deployment system is greatly improved, and it is suitable for the deployment of multiple types of service nodes in the cluster. At the same time, the space occupied by the image file on the image server node is significantly reduced.

D. In the transmission process of the operating system image, two network transmission modes of reliable multicast and unicast are adopted. Using the reliable multicast model to transmit the operating system image, the time is short and the efficiency is high, but it is required that the target nodes to be deployed should define the same operating system image; using the unicast mode to transmit the operating system image, different operations can be installed on multiple target nodes at the same time System image, but the deployment process takes a long time. The corresponding relationship between the operating system image and the node is determined by the administrator according to the application arrangement of the target node.

E. The installation and configuration of the application service software of the cluster target node and the configuration of the operating system identification information are automated, and the target node does not require manual intervention at all.

Description of drawings

Fig. 1 is the structural representation that the method of the present invention adopts. Among them, "DHCP SERVER" stands for "DHCP server", "TFTP SERVER" stands for "TFTP server", "Deploy SERVER" stands for "deployment server", and "Config SERVER" stands for "configuration server". One node in the figure is used as a deployment server node, providing DHCP service, TFTP service, deployment service and configuration service. The target node is connected to the deployment server node through a switch, downloads the image from the image server node and completes the installation.

Fig. 2 is a schematic diagram of main steps of cluster node deployment using the method of the present invention. The sequence of arrows describes the main steps of cluster deployment using this method.

FIG. 3 is a schematic diagram of a detailed process of booting a target node through PXE using the method of the present invention. It mainly describes the process of starting a target node from the network card through the PXE protocol, and then running the deployment agent in memory. where "bootstrap" exactly means bootstrap.

Detailed ways

Referring to Fig. 1, Fig. 1 is the structural representation that the method of the present invention adopts. Among them, "DHCP SERVER" stands for "DHCP server", "TFTP SERVER" stands for "TFTP server", "DeploySERVER" stands for "deployment server", and "Config SERVER" stands for "configuration server". One node in the figure is used as a deployment server node, providing DHCP service, TFTP service, deployment service and configuration service. The target node is connected to the deployment server node through a switch, downloads the image from the image server node and completes the installation.

The present invention adopts the client/server system structure, the installation mode based on the disk image, the deployment principle of separating the operating system image and the application service image, adopts the reliable multicast and unicast network transmission mode, and uses the network file system NFS and pre-boot execution Environment PXE protocol, dynamic host configuration protocol DHCP, simple file transfer protocol TFTP, WOL, etc. to realize the rapid and automatic deployment of the cluster server system.

In the architecture of this method, one node needs to be configured as the deployment server node, and one or more nodes need to be configured as the image server node, or the image server node and the deployment server node share one node.

The deployment server node, as the server in the deployment architecture, provides image acquisition, image transmission, node configuration, DHCP and TFTP services by installing the deployment service software package on it; it is responsible for image acquisition, image management, and target node operating system installation And the installation and configuration of various application service software required after the installation is completed. The target node is the client in the deployment architecture. After the network card is started, the deployment agent is obtained by communicating with the deployment server node. The deployment agent deploys the operating system image in a reliable multicast or unicast transmission mode according to the administrator's preset settings. to the local hard disk of the target node. The deployment agent is a service package containing an embedded operating system that runs in memory on the target nodes of the cluster. The deployment agent includes the target hardware automatic detection program, the deployment implementation program and the tools for operating the hard disk; the target hardware automatic detection program is responsible for automatically obtaining information on hardware components (boards and chipsets, etc.) and loading appropriate drivers based on relevant information , to ensure that the operating system can run on different hardware platforms. The deployment implementation program uses reliable multicast or unicast transmission mode to download and install the operating system image to the local hard disk of the target node according to the administrator's preset. The tools for operating the hard disk mainly include tools such as disk copy and disk partition.

The startup process of the target node complies with Intel's pre-boot execution environment protocol PXE; according to the PXE protocol, the deployment server node needs to provide Dynamic Host Configuration Protocol DHCP service and Simple File Transfer Protocol TFTP service.

Figure 3 mainly describes the process of a target node booting from the network card through the PXE protocol, and then running the deployment agent in memory. where "bootstrap" exactly means bootstrap.

When this method is actually implemented, both the deployment server node and the target node can use ordinary commercial computers, but the deployment server node must have input and output devices such as a keyboard, mouse, optical drive, floppy drive, and monitor, and the target node does not need these inputs. output device. The network card of the target node must be able to support PXE boot, so as to realize the full automation of the target node installation process. At present, general commercial computers can support the PXE boot function.

The target node, the deployment server node and the image server node are all connected by Ethernet, using a star network topology model, and all target nodes must be directly connected to the deployment server node through a switch or hub. The selected Ethernet switch must support the multicast function, otherwise, only the unicast transmission mode can be selected.

In this approach, the deployment agent is a service package containing an embedded operating system running in memory on the target node. The deployment agent includes target hardware automatic detection program, deployment implementation program and tools for operating the hard disk. At the same time, the deployment agent must support multicast and unicast transmission modes. When implementing the deployment agent, the deployment agent can be obtained by customizing the Linux operating system.

Fig. 2 is a schematic diagram of main steps of cluster node deployment using the method of the present invention. The sequence of arrows describes the flow of main steps of cluster deployment according to this method. Specific steps are as follows:

Step 1: Get the OS Image and App Service Image

1) Get the OS image

The process of obtaining the operating system image is similar to the working process of taking pictures, and the specific steps are as follows:

(1) Manually install the template node. As the source node of the image, at least the operating system and the configuration agent of the target node need to be installed on the source node (for configuring the node after the target node operating system image is deployed).

(2) After ensuring that the source node starts from the network card, the deployment server node controls the source node to download an image acquisition agent through PXE. The image acquisition agent is a service software package containing an embedded operating system, including tools for operating the hard disk . After the source node obtains the agent from the memory boot image, use the disk tool on the image acquisition agent to capture the disk image and store it as an operating system image file on the image server node, because the operating system image obtained from the source node will be completely cloned to the target node Therefore, this method is suitable for the case where the hardware devices of the target node and the source node are isomorphic.

2) Get the app service image

When obtaining the application service image, it is necessary to upload the application service software installation package (rpm, tar or gzip format) to the specified directory of the image server node, and create the corresponding application service description file as the application service image file.

Step 2: Define the target node

Bind the MAC address of the network card of the target node with the operating system image and the application service image, and determine the corresponding relationship between each image and the target node. At the same time, identify information such as an IP address and a host name of the target node are defined. The identification information such as IP address and host name here refers to the IP address and host name of each target node expected by the management personnel after the deployment of the target node is completed. One image file (operating system image or application service image) can correspond to multiple target nodes; one target node can only bind one operating system image file, but can bind multiple application service image files.

Step 3: Install the target node operating system

When installing the target node, you first need to prepare the operating system image and application service image on the image server node, and then you need to start the DHCP server, TFTP server, deployment server, configuration server, etc. of the deployment server node. The DHCP server provides the target node with a dynamically allocated IP address, specifies the location of the TFTP server, and notifies the target node to download the startup boot file from the location specified by the TFTP server.

The TFTP server provides services for the target node to download and deploy the agent program, and the boot boot file, startup configuration file, Linux kernel and root file system are stored in the root directory of the TFTP server. The target node downloads these files into the local memory and loads them for execution. After the target node starts the deployment agent from the memory, the deployment implementation program in the deployment agent sends a request to deploy the operating system image to the deployment server on the deployment server node, and the deployment server responds to the deployment according to the operating system image and image transfer mode type set by the administrator. To implement the request of the program, if the reliable multicast transmission mode is used, the deployment server will notify the target node to start the reliable multicast receiving program; if the unicast transmission mode is used, the deployment server will provide the storage location of the operating system image, and the deployment agent will In response, the operating system image is deployed to the hard disk of the target node using a corresponding transfer mode.

Step 4: Install the target node application service

After the target node completes the operating system installation, the deployment server node controls the target node to boot from the local hard disk through PXE, and the configuration agent contained in the deployed operating system image will communicate with the deployment server node to obtain the configuration information defined by the administrator for the target node , according to the obtained configuration information, download the corresponding application service software from the image server node and install and configure it.

Step 5: Configure the operating system information of the target node

The cluster system is a loosely coupled system. The nodes in the cluster are relatively independent. In order for the nodes in the cluster to cooperate with each other to provide external services, the identification information of the nodes must be configured. After installing and configuring the application service software of the target node, the configuration agent will configure the operating system information of the target node according to the administrator's definition, including the configuration of identification information such as host name and IP address.

After configuration, the target node is a cluster node that can be used immediately.

Step 6: Test: After the target node of the software system to be installed in the cluster system is installed, run the test program to test whether the installation and configuration of the operating system and application services are correct.

Claims (8)

1. the automated quick deployment method of an aggregated server system node, it is characterized in that, this method adopts client/server architecture, disposes the destination node of deployment server node, switch, one or more image server node, source node and software systems to be installed in this architecture;

All connect by Ethernet between destination node, deployment server node and the image server node, adopt star-like network topology model, all destination nodes will directly be communicated with the deployment server node by switch or hub; Wherein, the deployment server node is used to provide DHCP service, simple file transfer protocol service, deployment services and configuration service as the server end in the architecture;

When destination node is disposed, at first, destination node is started from network interface card, download the deployment agency by DHCP agreement and TFTP agreement from the deployment server node, the agency is disposed in operation in internal memory, dispose the agency by with the deployment server node communication after, finish operating system installation to this destination node; Then, deployment server node control destination node starts from local hard drive, and the operating system that operation was disposed is just now finished the installation of application service reflection and the configuration of operation system information by Configuration Agent; After configuration, this destination node is exactly a clustered node that can use immediately.

2. the method for claim 1 is characterized in that, the image server node adopts operation system image to separate deployment with the application service reflection based on the mounting means of disk image, wherein:

Operation system image is an image file that comprises the operating system that need install on destination node, by disposing the agency obtains source node by the disk image mode operating system copy generation;

The application service reflection is the image file of the application service software that need install on destination node, uploads and defines relevant XML description document by the keeper and generate; Two kinds of image files all are stored on the image server node, and the keeper is before destination node is installed, and according to the concrete application target dynamic combined reflection of destination node, destination node is by the image file on the NFS file system access image server node.

3. the method for claim 1 is characterized in that, when deployment server node during to destination node deployment operation system image, adopts reliable multicast or unicast networks transmission mode; Use NFS NFS, Pre-Boot eXecution Environment PXE agreement, dynamic host configuration protocol DHCP, simple file transfer protocol TFTP, WOL WOL to realize the rapid automatized deployment of group system.

4. method as claimed in claim 3, it is characterized in that described employing reliable multicast transmission mode is once disposed the destination node of one group of same operation system image, this pattern is applicable to that more the clustered node number is many, and the deployment of the less group system of OS Type to be installed.

5. method as claimed in claim 3, it is characterized in that described employing unicast transmission mode can be disposed many destination nodes with different operating system images simultaneously, be applicable to that more the clustered node number is few, and the deployment of the more group system of node OS Type to be installed.

6. method as claimed in claim 3, it is characterized in that, before destination node down load application service reflection, the image server node at first adopts the NFS file system that the application service reflection is shared in the network, download the application service reflection that needs by destination node by the communication pattern of clean culture then, finish the installation and the configuration of application service software.

7. the method for claim 1 is characterized in that, described image server node and deployment server node can share a computer node.

8. the method for claim 1, it is characterized in that, described deployment agency is a service software that comprises embedded OS that moves in target node memory, and this software comprises the automatic locator(-ter) of destination node hardware, the instrument of disposing implementation procedure and hard disk being operated;

The automatic locator(-ter) of destination node hardware is responsible for obtaining automatically hardware component information and is loaded suitable driver according to relevant information, guarantees can move this operating system on different destination node hardware platforms;

The deployment implementation procedure preestablishes according to the keeper's, adopts the transmission mode of reliable multicast or clean culture that operation system image is downloaded and is installed on the local hard drive of destination node;

The instrument that hard disk is operated mainly comprises diskcopy, disk partition instrument.