Host node state determination method and device, host node and storage medium

Abstract

The invention relates to the technical field of a host cluster, and provides a host node state determination method and device, a host node and a storage medium. The method comprises the steps of: acquiring a current detection period node heartbeat record of a second host node and a current detection period mounting signature of the second host node from a storage device according to a preset detection period; and according to consistency of the current detection period node heartbeat record of the second host node and on the basis of a mounting state indicated by the current detection periodmounting signature, determining a state of the second host node. According to the invention, the host node which is really in a dead state is determined by the node heartbeat record and the mounting signature, and the preset stable period is prolonged to implement accurate determination on the state of the host node so as to ensure that when each host node mounts a file system, states of member host nodes, which are sensed by the host node, are consistent, thereby avoiding the fatal problems of cluster disorder, server breakdown, even storage data loss and the like which are caused by inconsistency of the states.

Description

technical field [0001] The present invention relates to the technical field of host clusters, in particular to a method and device for determining the state of a host node, a host node and a storage medium. Background technique [0002] Multiple host nodes in the same host cluster can mount the same shared file system at the same time. When any host node in the host cluster operates on a specified file in the mounted file system, the mount includes the specified Other host nodes of the file can perceive the operation result of this operation. When the host node mounts the file system, it will first periodically write the node heartbeat record of the host node to the storage device. When the number of successful writes reaches the preset threshold, the host node is considered to be stable. At the same time, The host node will also periodically read the node heartbeat records of other host nodes in the host cluster except the host node from the storage device. If the values ​...

AI Technical Summary

Problems solved by technology

When the host node mounts the file system, it will first periodically write the node heartbeat record of the host node to the storage device. When the number of successful writes reaches the preset threshold, the host node is considered to be stable. At the same time, The host node will also periodically read the node heartbeat records of other host nodes in the host cluster except the host node from the storage device. If the values ​​of the node heartbeat records read in two cycles are different, the The host node is in the active state. If the values ​​of the node heartbeat records read in two cycles are the same, the host node is considered to be in the dead state. However, if the host node has a short-term storage link failure or other abnormalities, simply Judging that the host node is in the dead state based on the difference in the node heartbeat records of the two cycles will cause the host node that mounts the file system to think that the host node is dead and will not send a request to join the host cluster to the host node. When other host nodes are mounting At the same time, it is found that the host node is in the active state, so it sends a request to join the host cluster to the host node, and finally makes different host nodes in the entire host cluster perceive that the states of the member host nodes in the host cluster are inconsistent, which eventually leads to cluster chaos, Fatal problems such as server downtime or even storage data loss

Images