CN110750369B - A distributed node management method and system - Google Patents

Abstract

The present invention provides a distributed node management method and system. The method includes: establishing a secure communication channel between a management server and a working node, and the management server sends task information to the working node through the secure communication channel; the The working node determines whether an update operation needs to be started according to the task information; if so, the working node establishes a message queue with the message server, and sends an update request to the management server through the message server; the management server receives and process the update request, and send the update data to the worker node; the worker node periodically sends its node information to the management server through the message queue, and the management server receives and updates the worker node node information. By adopting the solution of the present invention, the technical effect of taking into account the controllability, flexibility and real-time performance of node management in a distributed system cluster is achieved.

Description

A distributed node management method and system

technical field

The invention belongs to the field of distributed system cluster management, in particular to a distributed node management method and system.

Background technique

At present, distributed detection technology is generally adopted in the field of Internet information detection. Through distributed deployment of working nodes, the distribution scheduling server comprehensively considers the number of tasks, computing resources of working nodes and network resources, and dispatches working nodes to perform Internet information detection tasks in a distributed manner. Node management includes the installation and configuration of worker nodes, initialization, plug-in deployment and plug-in update, node status control, and node resource management. The efficiency of node management directly affects the operation efficiency of the whole system. Therefore, node management plays a very important role in such systems. Realizing the automatic management of nodes can greatly reduce the workload that needs to be done manually in the process of system deployment and management, and is of great significance for improving the scalability of the system and the scale of detection nodes.

Generally speaking, node management usually includes three basic modes: active management, passive management and exception management. Active management means that the management node sends management instructions to the worker nodes according to certain strategies, such as polling, random, etc., to complete the functions of worker node deployment, initialization, and collection of node status. Passive management means that the management node passively receives messages such as requests and node status reported by the working nodes, and takes corresponding management actions according to the messages. Passive management has high real-time performance, but it is difficult to guarantee reliability when the working node is abnormal. Exception management refers to the restart, retry, update and other management operations taken when a worker node fails in computing, storage, network, etc. or during the operation of a plug-in. Exception management can be implemented in different ways according to specific exceptions.

To sum up, the above-mentioned prior art has the following shortcomings: active management is suitable for scenarios such as node installation, plug-in deployment, initialization, etc., but is not suitable for management operations of individual nodes. Active management is highly operable and easy to control, but not flexible enough; passive management has high real-time performance, but it is difficult to guarantee reliability when working nodes are abnormal; exception management can only be used for abnormal situations in working nodes, and in other cases , it is not applicable.

SUMMARY OF THE INVENTION

In order to solve the above technical problems, the present invention proposes a distributed node management method, which includes the following steps:

Step 1: establish a secure communication channel between the management server and the working node, and the management server sends task information to the working node through the secure communication channel;

Step 2: the working node determines whether an update operation needs to be started according to the task information; if so, the working node establishes a message queue with a message server, and sends an update request to the management server through the message server;

Step 3: the management server receives and processes the update request, and sends the update data to the working node;

Step 4: The working node periodically sends its node information to the management server through the message queue, and the management server receives and updates the node information of the working node.

Further, on the basis of the above technical solution, the step 1 includes:

The management server actively builds an SSH channel to the working node. After the SSH channel is established, the management server sends the task information to the working node through the SSH channel, and controls the working node to complete the corresponding task information. function or operation.

Further, on the basis of the above technical solution, the task information includes plug-ins and version information thereof required to perform the task.

Further, on the basis of the above technical solution, the step 2 includes:

The work node compares the received plug-in and its version information with the local plug-in information of the work node to determine whether the update operation needs to be started; if not, the update operation needs to be started.

Further, on the basis of the above technical solution, the update operation includes:

The worker node establishes an RPC message queue with the message server, sends an update request to the message server through the RPC message queue, and then the message server sends the update request to the management server.

Further, on the basis of the above technical solution, the RPC message queue is provided by RabbitMQ.

Further, on the basis of the above technical solution, the step 3 includes:

The management server sends the latest version of the plug-in to the worker node through the message queue.

Further, on the basis of the above technical solution, the step 4 includes:

The worker node periodically sends its node information to the RPC message queue, and reports the resource and status information of the worker node to the management server through the message server. After the management server receives the node information, Update the node information of the working node saved by it.

Further, on the basis of the above technical solution, the node information in the step 4 includes node resource occupation and node status.

Further, on the basis of the above technical solution, the periodicity in the step 4 is implemented according to a preset timed heartbeat interval.

On the other hand, the present invention also provides a distributed node management system, including a processor and a memory, the memory has a medium storing program codes, when the processor reads the program codes stored in the medium, The system can execute the method described in the above technical solution.

By adopting the method and management system of the present invention, the problems existing in the existing node management methods that controllability, flexibility and real-time performance cannot be achieved simultaneously are solved. Since the present invention adopts a management method combining active and passive, it achieves at least the following technical effects: (1) realizes automatic batch initialization of working nodes, improves the deployment efficiency of working nodes, and has good controllability; (2) realizes detection plug-ins The automatic real-time update ensures that the working nodes can use the latest version of the plug-in to perform detection tasks, which improves the real-time performance of node management; (3) realizes the timing management and control of node resources and status, and realizes the timing collection and management of node status. , with a high degree of flexibility.

Description of drawings

FIG. 1 is a schematic flowchart of a distributed node management method proposed by the present invention

2 is a schematic diagram of automatic deployment of nodes based on SSH channels in the distributed node management method proposed by the present invention;

3 is a schematic diagram of the automatic update of the node plug-in based on the RPC message queue in the distributed node management method proposed by the present invention;

FIG. 4 is a schematic diagram of node resource and state timing management based on RPC message queue in the distributed node management method proposed by the present invention.

Detailed ways

The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

In order to better illustrate the present invention and facilitate the understanding of the technical solutions of the present invention, typical but non-limiting examples of the present invention are as follows: It should be noted here that the embodiments listed in the description of the present invention are only for the convenience of illustrating the problem. The given exemplary implementation method should not be construed as the only correct embodiment of the present invention, nor should it be construed as a limiting description of the protection scope of the present invention.

The goal of distributed node management is to minimize manual participation and improve the deployment efficiency of worker nodes; when the system detects an update of the plug-in version, the worker node can update the probe plug-in in time and perform the detection task with the latest version of the probe plug-in; the management node can Periodically collect worker node status and resource usage.

To this end, the present invention proposes a new technical solution that organically integrates SSH (Secure Shell Protocol) channel technology, RPC message queue, timing management and other technical means into node management. See Figure 1 for a schematic flow diagram, including:

Step 1: establish a secure communication channel between the management server and the working node, and the management server sends task information to the working node through the secure communication channel;

Step 2: the working node determines whether an update operation needs to be started according to the task information; if so, the working node establishes a message queue with a message server, and sends an update request to the management server through the message server;

Step 3: the management server receives and processes the update request, and sends the update data to the working node;

Step 4: The working node periodically sends its node information to the management server through the message queue, and the management server receives and updates the node information of the working node.

In order to facilitate the understanding of the above steps, each step of a preferred specific implementation manner of the technical solution proposed by the present invention is illustrated in the following.

Referring to FIG. 1, which is a preferred specific implementation of step 1 of the technical solution proposed by the present invention, the management server is based on the node automation management technology of the SSH (Secure Shell Protocol: Secure Shell Protocol) channel, using the SSH protocol, such as based on the libssh library (For details, see: http://www.libssh.org/) Build an SSH channel between the management server and the worker nodes. This SSH channel provides a channel for the management server to actively manage worker nodes. After the SSH channel is established, the management server can send commands to the working nodes, where the commands include task information, and control the working nodes to complete corresponding functions or operations.

In practical applications, this channel can be used to control worker nodes to download automatic initialization plug-ins, realize batch management of worker nodes, and improve worker node deployment efficiency.

Referring to Figure 2, the automatic deployment process of nodes based on SSH channel is as follows: (1) An SSH channel is established between the management server and the worker nodes (see Figure 1: (1) SSH channel is established); (2) The management server sends the SSH channel to the The worker node sends the command (see Figure 1: (2) Sending the command); (3) The worker node executes the command to complete a specific function (see Figure 1: (3) The command command is completed); (4) The worker node executes the command through the SSH channel The result is returned to the management server (see Figure 1: (4) return result).

The SSH channel-based node automation management technology provides a technical approach for the management server to actively manage the working nodes, and is mainly used in the initial deployment and debugging stages of the working nodes. During the operation of the system, due to the continuous improvement, update and perfection of the version of the detection plug-in, the version of the detection plug-in cached on the worker node may become outdated. It is of great significance to detect the change of plug-in version in real time and update the plug-in in time to ensure the quality and consistency of detection results. However, at this time, if you rely on the management server to check the plug-in versions of the worker nodes one by one through the SSH channel, there are the following problems: First, each task is required to be checked, resulting in a lot of unnecessary interactive traffic; second, it is easy to expose the management server IP address, causing The remote computer room blocks the IP address of the management server.

In order to avoid the above problems, see FIG. 3 , which is a preferred specific implementation of step 2 of the technical solution proposed by the present invention. The task information distributed to the worker nodes includes the plugins and version information required to execute the task. The worker nodes compare the received information with the local plugin information, and if they are inconsistent, start the plugin update. At this time, the worker nodes and the management Establish a remote call service RPC message queue between servers, such as: working nodes use RabbitMQ (an open source message broker software (also known as message-oriented middleware) developed by Rabbit that implements the Advanced Message Queuing Protocol (AMQP)) to provide RPC message queue, send the plug-in request to the message server, and then send the plug-in update request to the management server through the message server, the management server receives and processes the plug-in request, sends the latest version of the plug-in to the temporary message queue of the message server, and passes the message queue. Send the plug-in request result to the worker node to realize the automatic update function of the detection plug-in.

Node plug-ins are updated based on RPC message queues in a passive management and real-time update mode, as shown in Figure 3. The specific process is as follows: (1) During the task distribution process, the management server Manage Server specifies the plug-ins and The version is identified by the unique MD5 of the plug-in, and the task is distributed to a certain worker node Worker through the message server. (2) After receiving the task, the worker node Worker finds out whether there is a local plug-in that matches the execution of the task, and if not, starts the plug-in update process. (3) The worker node Worker puts the plug-in update request into the RPC message queue in the message server, and at the same time, establishes a dedicated temporary message queue for storing the plug-in request result returned by the management server. The worker node Worker continues to listen to this temporary queue until the plugin request result is retrieved. (4) The Manage Server side of the management server continuously monitors the RPC message queue of the message server MQ Server, retrieves the new messages in it, uniquely determines the updated plug-in requested by the worker node Worker through the plug-in MD5 value, and puts the latest version of the plug-in into the corresponding Temporary message queue. (5) The worker node Worker retrieves the plug-in request result from the temporary message queue created by itself, and realizes the real-time update of the plug-in version. At the same time, log off the corresponding temporary message queue.

Referring to FIG. 4 , it is a preferred specific implementation of step 3 of the technical solution proposed by the present invention. The worker node sends the node resource occupancy and node status to the RPC message queue according to the preset timing heartbeat interval, reports the node resource and status to the management server through the message server, and the management server receives and updates the corresponding node information. As a more preferred embodiment, the worker node Worker sends the node resource occupancy and node status to the RPC message queue in the form of a timed request according to a preset timed heartbeat interval, and the RPC message queue sends the above-mentioned timed request to the management through the message server MQServer. The server Manage Server reports node resources and status, and the management server receives and updates the corresponding node information. It can be seen that the difference between managing node resources and status based on RPC message queues and the automatic update of node plug-ins is that the automatic update of plug-ins is irregular, while using RPC message queues to manage node resources and status management is a regular and regular task. Therefore, based on the RPC message queue management node resources and status adopt passive management and regular update mode. Referring to Figure 3, the specific process is as follows: (1) The worker node Worker sends a timing request to the message server MQ Server by setting a timing heartbeat (for example, 30 seconds), and regularly sends the node resource occupancy and node status information to the RPC message queue , and at the same time, a dedicated temporary queue Qt is established to store the timing request results returned by the management server. The worker node Worker continues to monitor the temporary queue Qt until the timing request result is retrieved. (2) The Manager Server side of the management server continuously monitors the RPC message queue of the message server MQ Server, retrieves new messages, that is, new timing requests, parses the retrieved new messages, and updates the node status and resource occupancy. At the same time, according to the user operation, the command that needs to be executed by the worker node Worker is put into the corresponding temporary message queue as the result of the timing request. (3) The worker node Worker retrieves the message in the timing request result in the temporary message queue Qt, executes a specific command, and then waits for the next heartbeat to report the node status and resource occupancy.

For those skilled in the art, it is obvious that the embodiments of the present invention are not limited to the details of the above-mentioned exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or essential features of the embodiments of the present invention example. Accordingly, the embodiments are to be considered in all respects as exemplary and not restrictive, the scope of the embodiments of the present invention being defined by the appended claims rather than the foregoing description, and are therefore intended to fall within the scope of All changes within the meaning and scope of equivalents of the claims are included in the embodiments of the present invention. Any reference signs in the claims shall not be construed as limiting the involved claim. Furthermore, it is clear that the word "comprising" does not exclude other units or steps and the singular does not exclude the plural. Multiple units, modules or means recited in the system, device or terminal claims can also be implemented by the same unit, module or means by software or hardware. The terms first, second, etc. are used to denote names and do not denote any particular order.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention and not to limit them. Although the embodiments of the present invention have been described in detail with reference to the above preferred embodiments, those of ordinary skill in the art should Modifications or equivalent replacements to the technical solutions of the embodiments of the present invention should not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (7)

1. A distributed node management method is characterized in that the following steps are executed among a management server, a message server and a working node:

step 1: establishing a secure communication channel between a management server and a working node, wherein the management server sends task information to the working node through the secure communication channel, the management server actively establishes an SSH channel to the working node, after the SSH channel is established, the management server sends the task information to the working node through the SSH channel to control the working node to complete corresponding functions or operations, and the task information comprises plug-ins and version information thereof required by task execution;

step 2: the working node judges whether to start updating operation or not according to the task information; if so, establishing a message queue by the working node and a message server, and sending an updating request to the management server through the message server, wherein the working node compares the received required plug-in and version information thereof with local plug-in information of the working node to judge whether the updating operation needs to be started or not; if not, the updating operation needs to be started; the message queue established by the working node and the message server is an RPC message queue, an updating request is sent to the message server through the RPC message queue, and then the message server sends the updating request to the management server;

and step 3: the management server receives and processes the updating request and sends updating data to the working node;

and 4, step 4: and the working node periodically sends the node information to the management server through the message queue, and the management server receives and updates the node information of the working node.

2. The method of claim 1, wherein the RPC message queue is provided using a RabbitMQ.

3. The method of claim 2, wherein said step 3 comprises:

and the management server sends the latest version of plug-in to the working node through the message queue.

4. The method of claim 3, wherein said step 4 comprises:

the working node sends the node information of the working node to the RPC message queue periodically, the resource and state information of the working node is reported to the management server through the message server, and the management server updates the node information of the working node stored by the management server after receiving the node information.

5. The method of claim 4, wherein the node information in step 4 includes node resource occupation and node status.

6. The method according to claim 5, wherein the periodicity in step 4 is implemented according to a preset timed heartbeat interval.

7. A distributed node management apparatus comprising a processor and a memory, the memory having a medium with program code stored thereon, the apparatus being capable of performing the method of any one of claims 1 to 6 when the processor reads the program code stored on the medium.

Images