A method and device for detecting and processing cluster split brain

Abstract

The invention discloses a method of detecting and processing brain split in a cluster, the cluster comprises a first node and a second node, the method further comprises the following steps that: the first node sends a first ARP (Address Resolution Protocol) detection message generated by the first node to the second node and receives a second ARP detection message from the second node; when the first node determines that the first node and the second node belong to one cluster according to the detection domains of the first ARP detection message and the second ARP detection message, the first node judges whether the brain split occurs; and when the brain split is judged, the first node executes closing operation or normal operation according to the detection domains of the first ARP detection message and the second ARP detection message. The invention further discloses a device of detecting and processing the brain split in the cluster.

Description

technical field [0001] The invention relates to cluster virtualization technology, in particular to a method and device for detecting and processing cluster split-brain. Background technique [0002] With the increasing business of data centers and the continuous improvement of user requirements, the scale and functions of data centers are becoming more and more complex, and the management difficulty is also becoming more and more difficult. In this context, integrating data centers, reducing data center management costs, and fully tapping existing resource capabilities to meet higher business needs have become important tasks for enterprise data centers. Among them, virtualizing data center resources has become an important trend in data center integration. [0003] Data center virtualization technology mainly includes three aspects: network virtualization, storage virtualization and server virtualization, the most important of which is server virtualization. Such as fig...

AI Technical Summary

Problems solved by technology

But when there are only two nodes in the HA cluster, each node has one vote and cannot be compared; for this case, the existing solution is to introduce third-party equipment, but third-party equipment has potential unreliability factors

[0009] In the data center virtualization environment, for the above situation where there are only two nodes in the HA cluster and a split brain occurs, according to the protection mechanism of the HA cluster, the system will try to start the same virtual machine on the two nodes in the cluster respectively. , two copies of the same virtual machine run at the same time, which will lead to abnormal business of the virtual machine. In addition, two virtual machines reading and writing the same image file may cause damage to the virtual machine image file

Images