CN103647830A - Dynamic management method for multilevel configuration files in cluster management system - Google Patents

Abstract

The present invention provides a dynamic management method for multi-level configuration files in a cluster management system. When the cluster is started, the multi-level tree is initialized according to the information of the cluster configuration files; during the operation of the cluster, the configuration files are dynamically modified and managed according to user needs; nodes When a fault occurs and then recovers, the dynamically modified configuration file is protected against faults. Compared with the existing technology, the dynamic management method of the multi-level configuration file in the cluster management system ensures the sustainable operation of the business, changes the traditional modification based on the cluster or group, and reduces the coupling degree between nodes. Realize multi-node multi-task simultaneous operation and improve system performance.

Description

A dynamic management method of multi-level configuration files in a cluster management system

technical field

The invention relates to the technical field of computer applications, in particular to a dynamic management method for multi-level configuration files in a cluster management system.

Background technique

The high-availability cluster management system includes the management of multiple business groups, and each group has multiple resources and monitoring resources. The cluster configuration file records cluster, group, resource, monitoring resource, server and cluster management policy information, which is the only basis for cluster management. Traditional configuration file management is static management, which requires that the configuration file cannot be changed after the cluster is started, unless the cluster is stopped, the configuration file is resynchronized, and then the cluster is started. This not only interrupts business operations and cannot provide sustainable protection, but also increases a lot of human operations and is prone to errors.

Contents of the invention

The technical task of the invention is to solve the deficiencies of the prior art and provide a dynamic management method for multi-level configuration files in a cluster management system.

The technical solution of the present invention is realized in the following manner, the dynamic management method of the multi-level configuration file in the cluster management system, its specific dynamic management process is:

Set the dynamic management module of the multi-level configuration file: the initialization module, the dynamic management module and the fault-tolerant module. The three read the cluster configuration file, communicate and negotiate for the configuration change, and cooperate to complete the generation and change of the configuration.

When the cluster is started, the initialization module reads the configuration file information and builds a multi-level cluster tree, directly completing the dynamic management configuration of the cluster configuration file.

During cluster operation, the dynamic management module is used to dynamically manage the cluster tree: dynamically modify the information to traverse the tree according to user needs, modify the configuration file from the smallest granularity, and synchronize it to other nodes in the cluster.

When a node fails and then recovers, the fault tolerance module is used for file-level disaster recovery: when a node fails and recovers, the recovery node sends a message to the running node to obtain the latest configuration file, ensuring the consistency of the configuration file information of each node in the cluster, and completing the dynamic Modified configuration files are protected for fault tolerance.

The initialization module represents clusters, groups, resources and monitoring resource nodes in the high-availability cluster management system as a "multi-level" tree, which is a logical graph structure based on the GOM model.

The multi-level tree model divides a tree into several levels, and the subtree depth of each level is not greater than 2. There are two types of nodes in the tree, one is extended nodes, corresponding to resources and monitoring resources in the cluster; the other is are unscaled nodes, corresponding to clusters and groups in the cluster.

Each of the above nodes contains a set of operation primitives and attribute configuration: the operation primitives express the dynamic change operations for nodes or subtrees, these operations include adding Add, deleting Delete, upgrading Upgrade, downgrading Degrade, replacing Replace and copying Clone ;Property configuration describes the attributes of the cluster, group, resource and monitoring resource node, corresponding to the information of the current cluster configuration file, and describes the health value of whether the node can communicate with other nodes.

The beneficial effect that the present invention produces compared with prior art is:

The dynamic management method of multi-level configuration files in a cluster management system of the present invention can realize the dynamic management of cluster configuration files, and through this method, the groups and resource nodes in the cluster configuration files are divided into layers, which can be used in smaller Granularity modifies multiple resources and resource attributes at the same time, enhances the dynamic and intelligent management of clusters, and changes the traditional modification in units of clusters or groups. The cluster is constructed with a multi-level tree structure, which reduces the coupling degree between nodes, thereby realizing the simultaneous operation of multiple nodes and multiple tasks, and improving the performance of the system. It is practical and easy to promote.

Description of drawings

Accompanying drawing 1 is the schematic diagram of the cluster multi-level tree structure of the present invention.

Accompanying drawing 2 is a schematic diagram of the configuration file dynamic management scheduling of the present invention.

Figure 3 is a schematic diagram of adding group resources in the present invention.

Detailed ways

A method for dynamically managing multi-level configuration files in a cluster management system of the present invention will be described in detail below in conjunction with the accompanying drawings.

The present invention provides a dynamic management method for multi-level configuration files in a cluster management system, which can realize dynamic management of cluster configuration files, and divide the groups and resource nodes in the cluster configuration files into layers according to the method, which can be implemented in a smaller Simultaneously modify multiple resources and resource attributes at the granularity to enhance the dynamic and intelligent management of clusters. The specific dynamic management process is as follows:

1. Set up the dynamic management module of the multi-level configuration file: the initialization module, the dynamic management module and the fault-tolerant module. The three read the cluster configuration file, communicate and negotiate for the configuration change, and cooperate to complete the generation and change of the configuration.

2. The cluster starts, the initialization module reads the configuration file information and builds a multi-level cluster tree, and directly completes the dynamic management configuration of the cluster configuration file.

In the above technical solution, the communication primitives based on the graph structure are used between nodes to avoid direct code calls, thereby achieving loose coupling of the system.

The communication content between nodes adopts the form of JSON, which can simplify the analysis of node stored information and the difficulty of analyzing the returned results; at the same time, the use of a fixed JSON format can shield the implementation details from the outside, enhance system security and reduce task triggers Difficulty of realization.

The configuration function based on the graph structure allows the user to choose whether to use the configuration file for dynamic configuration or directly perform pre-arranged dynamic configuration in the program, which increases the system flexibility.

3. During cluster operation, the dynamic management module is used to dynamically manage the cluster tree: dynamically modify the information traversal tree according to user needs, modify the configuration file from the smallest granularity, and synchronize it to other nodes in the cluster.

In this technical solution, the group and resource information can be dynamically modified without interrupting the business running time; the modification with the smallest granularity reduces the interaction with other nodes and improves the system performance; it supports multiple dynamic modification operations and improves the flexibility of the system.

4. When a node fails and then recovers, the fault tolerance module is used for file-level disaster recovery: when a node fails and recovers, the recovery node sends a message to the running node to obtain the latest configuration file to ensure the consistency of the configuration file information of each node in the cluster. Fault-tolerant protection for dynamically modified configuration files.

Further, the above steps are further described in conjunction with accompanying drawings 1 to 3 as follows:

The initialization module is used to read the configuration file to build a multi-level cluster tree when the cluster starts, and directly complete the dynamic management configuration of the cluster configuration file.

The initialization module represents the cluster, group, resource and monitoring resource nodes in the high-availability cluster management system as a "multi-level" tree, which is a logical graph structure based on the GOM (Graph Oriented Model) model. This logical graph structure not only enables users to use various communication primitives on the graph to communicate with adjacent layers, but also provides a clear framework for dynamic configuration.

The multi-level tree model divides a tree into multiple levels, and the sub-tree depth of each level is not greater than 2. To define a sub-tree, it is only necessary to determine its root node and child nodes. There are two types of nodes in the tree, one is the expanded node, the other is the unexpanded node, and the unexpanded node itself is a tree. As shown in Figure 1, the cluster tree Tree1 is divided into a multi-level tree Tree2, in which the expanded nodes correspond to the resources and monitoring resources in the cluster, and the unexpanded nodes correspond to the clusters and groups in the cluster.

Each node contains a set of operation primitives and attribute configuration.

Operation primitives express dynamic change operations for nodes or subtrees, including: Add, Delete, Upgrade, Degrade, Replace, and Clone. These graph-based configuration functions allow users to choose whether to use configuration files for dynamic configuration or to perform pre-arranged dynamic configuration directly in the program, thus increasing system flexibility. Because it is possible that in some cases, the conditions of the dynamic configuration of the node during operation remain unchanged, and there is no need to describe it in the configuration file for future changes. Additionally, allowing the use of system configuration functions in nodes makes sense for adding new nodes non-deterministically. After the new node completes the necessary initialization work, it attaches itself to a node in the tree structure through the AddNode function, and can be seamlessly embedded in the system without other work. The addition of a new node will not affect the operation of its parent node, as long as the parent node fully adopts the communication primitives based on the graph structure.

Each node can only communicate with its parent node, child nodes and sibling nodes, and cannot communicate across layers. The purpose of this is to strengthen system control, and when dynamic configuration occurs, it will affect the normal operation of other parts of the system as little as possible.

On the one hand, attribute configuration describes the attributes of clusters, groups, resources, and monitoring resource nodes, which are in one-to-one correspondence with the current cluster configuration file information. On the other hand, it describes the communicable state of the node, that is, the health value of whether the node can communicate with other nodes. The cluster attributes contained in the nodes are shown in Table 1.

The dynamic management module is used to dynamically manage the cluster tree. The dynamic management process is closely related to the operation primitives of the nodes. It is an operation on the tree and will not be described in detail here.

The fault tolerance module is used for file-level disaster recovery. For example, in a two-machine cluster, after the peer end fails, the cluster configuration file is dynamically modified, and then the peer end recovers, but the configuration file before the fault occurs runs on it, resulting in inconsistency between the two ends. The fault-tolerant module will make the recovery end send a recovery message to the normal operation node to obtain the latest configuration file information.

Each module is required to modify the configuration file in an exclusive way to maintain the consistency of the cluster configuration file.

In a typical environment, after the user configures the cluster, synchronizes the configuration file, and starts the cluster, the initialization module first builds a multi-level tree based on the information in the cluster configuration file. The root node of the tree stores the cluster information, and all non-child nodes below the root node The group information is stored, and the dependencies between groups are represented in a hierarchy, and the upper group nodes depend on their child group nodes. All child nodes in the multi-level tree store resources or monitor resource information, and the parent node of all child nodes must be a group node. The table below shows the cluster configuration file information stored by different types of nodes.

node name Node attribute information cluster Cluster name, cluster IP, subnet mask, network partition analysis, use heartbeat, master server information, heartbeat timeout information, alarm service, etc. Group Group name, start server, number of failovers, exclusive groups, group dependencies resource Resource type, resource name, start (stop) timeout, start (stop) failure retry times, start (stop) failure final action, resource-specific attributes Monitor resources Monitoring resource type, Monitoring resource name, Interval, Timeout, Retries, Waiting time to start monitoring, Target resource

As shown in Figure 1, the cluster configuration file is constructed into a cluster tree Tree1, which is divided into a multi-level tree Tree2 according to the level. The unexpanded nodes of each level are group resources, and the hierarchical relationship between groups represents the relationship between groups. dependencies. A group node must be a subtree whose children are either dependent groups or group resources or group monitoring resources. Each layer that is both an extension node and a child node must be a resource or monitor resource node.

When the cluster is running, if the user wants to modify the cluster configuration file, there is no need to stop the cluster first and then modify it. Only need to modify the cluster configuration file, click the "Apply" button, the cluster system will pass the information modified by the user to the root node of the cluster, read the modified information, the dynamic management module agent of the root node and the dynamic management module agent of each layer communicate and negotiate with each other , decide which layer to modify, and perform corresponding operations. Figure 3 shows the dynamic addition of group resource fipAdd, traverse the tree to find the group failover1 node, add fipAdd resource and resource information through the Addnode function, and synchronize the group resource information on each node of the cluster.

The above description is only an embodiment of the present invention, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims (4)

1. a dynamic management method of multi-level configuration file in a cluster management system, characterized in that its concrete dynamic management process is: Set the dynamic management module of the multi-level configuration file: the initialization module, the dynamic management module and the fault-tolerant module. The three read the cluster configuration file, communicate and negotiate for the configuration change, and cooperate to complete the generation and change of the configuration; The cluster starts, the initialization module reads the configuration file information and builds a multi-level cluster tree, and directly completes the dynamic management configuration of the cluster configuration file; During cluster operation, the dynamic management module is used to dynamically manage the cluster tree: dynamically modify the information traversal tree according to user needs, modify the configuration file from the smallest granularity, and synchronize it to other nodes in the cluster; When a node fails and then recovers, the fault tolerance module is used for file-level disaster recovery: when a node fails and recovers, the recovery node sends a message to the running node to obtain the latest configuration file, ensuring the consistency of the configuration file information of each node in the cluster, and completing the dynamic Modified configuration files are protected for fault tolerance. 2. the dynamic management method of multi-level configuration file in a kind of cluster management system according to claim 1, it is characterized in that: described initialization module expresses cluster, group, resource and monitoring resource node in high-availability cluster management system as "Multi-level" tree, which is a logical graph structure based on the GOM model. 3. the dynamic management method of multi-level configuration file in a kind of cluster management system according to claim 2, it is characterized in that: multi-level tree model divides a tree into several levels, and the subtree depth of each level is not more than 2 , there are two types of nodes in the tree, one is extended nodes, corresponding to resources and monitoring resources in the cluster; the other is unexpanded nodes, corresponding to clusters and groups in the cluster. 4. The dynamic management method of a multi-level configuration file in a cluster management system according to any one of claims 1 to 3, characterized in that: each of the above-mentioned nodes contains a group of operation primitives and attribute configuration: operation primitives The language describes the dynamic change operations for nodes or subtrees. These operations include adding Add, deleting Delete, upgrading Upgrade, downgrading Degrade, replacing Replace and copying Clone; attribute configuration describes the attributes of clusters, groups, resources, and monitoring resource nodes. One-to-one correspondence with the current cluster configuration file information, and at the same time describe the health value of whether the node can communicate with other nodes.

Images