Health self-checking method and system for each node of distributed storage cluster

Abstract

The invention discloses a health self-checking method and system for each node of a distributed storage cluster, which are applied to a main management node in the cluster. According to the invention, the upper limit of the number of the processes concurrently executing health self-inspection is controlled in a process pool mode, meanwhile, the recycled processes are reused to continue to execute the health self-inspection tasks of the remaining nodes, and system expenditure caused by frequent creation and destruction of the processes is avoided. Moreover, by setting the upper limit of the concurrent execution process of the process pool as the number of CPU cores of the main management node, the consumption of system resources is reduced while the speed of obtaining a node health self-check result is maximized, so that the resource consumption and the execution speed of process health self-check are optimally balanced, and the invention has good operability and expansibility under large-scale cluster nodes.

Description

technical field [0001] The invention relates to the field of distributed storage, in particular to a health self-check method and system for each node of a distributed storage cluster. Background technique [0002] At present, the health self-check of each node in a distributed storage cluster is implemented in a concurrent multi-process manner. Specifically, corresponding processes are started one by one according to the number of nodes in the cluster to obtain the health status of each node in the cluster. That is to say, How many nodes there are in the cluster, how many processes will be started at the same time. However, when the number of nodes in the cluster reaches hundreds or thousands or even more, the original concurrent multi-process mode starts too many processes, which will cause serious occupancy of system resources, and the single CPU (central processing unit) of the system ) can only process one process task at a time, and processes with more than the number...

AI Technical Summary

Problems solved by technology

However, when the number of nodes in the cluster reaches hundreds or thousands or even more, the original concurrent multi-process mode starts too many processes, which will cause serious occupancy of system resources, and the single CPU (central processing unit) of the system ) can only process one process task at a time, and processes with more than the number of CPUs can only be queued and processed according t

Images