Double-computer reinforcing method for high-performance job scheduling management node

Abstract

The invention discloses a double-computer reinforcing method for a high-performance job scheduling management node. The method is mainly applied in the field of high-performance calculation and is used for achieving dual reinforcement of a node level and a resource level by means of installation and configuration of Pacemaker, corosync and drbd software in terms of the job scheduling management node. According to the method, on one hand, a single-point fault problem of single-computer deployment is avoided, and on the other hand, job system resource monitoring can be achieved through the pacemaker software, and storage of a job scheduling system can be configured on two computers through the drbd software; compared with a heartbeat+NFS mode, namely an NFS shares storage of the job scheduling system, the method has dual redundant advantages, effectively guarantees operation reliability of the system and can effectively make up for defects of the traditional method.

Description

technical field [0001] The invention relates to the field of computers, in particular to a Pacemaker+corosync+drbd-based dual-computer reinforcement method for high-performance job scheduling management nodes. Background technique [0002] Currently, network-based computer technology promotes the development and wide application of cluster systems. Use a high-speed network to connect high-performance workstations or PCs into clusters according to a certain structure to achieve parallel computing. With only a small cost, you can get the performance of mainframes and parallel computers; however, with the increase in the application scale of high-performance computing clusters With continuous expansion, cluster management issues also follow. The job scheduling system is mainly responsible for receiving job requests submitted by users, and selecting appropriate computing resources to complete user jobs according to specific scheduling rules and user requirements for jobs. With...

AI Technical Summary

Problems solved by technology

[0003] However, the traditional job scheduling system often adopts a stand-alone method or uses heartbeat+NFS method for dual-machine reinforcement. Both methods have certain defects and loopholes, such as management If the node is deployed on a single machine, once the node fails, the job scheduling system of the entire high-performance cluster will stop working, and the jobs of the entire cluster cannot be scheduled reasonably and effectively, so the job operation will stagnate, seriously affecting the operating efficiency of the system; another example is using heartbeat +NFS mode is used for dual-machine reinforcement. Due to the design factors of the heartbeat software itself, resource-level monitoring of the job scheduling system cannot be implemented. Once the monitored resource fails, resource switching cannot be effectively performed, which will cause the entire cluster operation to fail. Effective scheduling seriously affects system operation efficiency

It can be seen that both traditional security hardening methods have fatal shortcomings, so how to more effectively harden the job scheduling system has become an urgent problem to be solved

Images