Method and device for realizing high availability of cross-node virtual machine cluster

Abstract

The invention discloses a method and a device for realizing high availability of a cross-node virtual machine cluster. The method comprises the following steps of establishing a cluster through a pacemaker resource management tool; obtaining network connection modes and virtual addresses of virtual machines; obtaining information of a physical network card use for providing a heartbeat link for the virtual machines on a host; obtaining a state of the cluster; and querying whether to carry out fence operation on the virtual machines on an opposite terminal host through calling a fence script on an opposite terminal node, if the number of failures of the fence operation exceeds a preset threshold, determining the physical network card of the host for causing the fissure and removing the virtual machines on the host for causing the fissure from the cluster. According to the method and the device for realizing the high availability of the cross-node virtual machine cluster, a problem of failure of the physical network card of the host or cluster fissure caused by failure of an external switch can be solved, the cluster state exception caused by an external factor is avoided, and the effective external service provision is guaranteed.

Description

technical field [0001] The invention belongs to the technical field of cluster communication, and in particular relates to a method and device for realizing high availability of a cross-node virtual machine cluster. Background technique [0002] In the prior art, the cluster solution is based on the high-availability features of virtual machine clusters implemented by pacemaker resource manager and its components corosync (providing heartbeat support), fence_virsh (providing fault recovery), and so on. The node sends a heartbeat signal. When a node's heartbeat is lost, other nodes perform fence operations on the node by starting fence script resources, including restarting and shutting down the node. [0003] Two or more nodes that communicate with each other in the cluster usually send heartbeat signals to ensure that the nodes are running normally. Usually, when a node fails (downtime, crash), it will stop the peer node. Sending a heartbeat signal, the peer node consider...

AI Technical Summary

Problems solved by technology

When the network cable of the host physical network card is pulled out due to human factors or other factors, the heartbeat of the nodes in the cluster is lost. In this case, when the heartbeat is lost, the nodes at both ends of the link cannot detect the heartbeat signal of the peer node. , the fence operation will continue to be performed on the peer node, but because the physical network cable of the node is pulled out or damaged, the fence operation on the nodes at both ends of the link will not succeed, resulting in a split-brain state of the cluster

When the cluster is in the split-brain state, the node in the state of losing heartbeat will continue to send fence operations to allow the peer node to perform fence operations, but the fence operations will not be executed successfully, and the result is that the state of the entire cluster is abnormal , unable to continue to provide services outside the cluster normally

Images