Method and system for calling distributed file system

Abstract

The invention discloses a method and a device for calling a distributed file system. One specific implementation way of the method comprises the following steps of: on the basis of a received session request of a user side, establishing a new session; sending a first-time locking request to an administrator of the distributed file system, wherein the first-time locking request is used for requesting to lock a lease directory; responding when a response to the first-time locking request is success in locking is received, and sending a request for establishing a new subdirectory in the lease directory to the administrator, wherein the subdirectory is named by the IP (Internet Protocol) address and the port of a new session; and responding when a response to the request for establishing the new subdirectory is success in establishing is received, and marking the new session as a main session. By use of the implementation way, the main session selection of the new session is realized, the main session is locked, robustness for calling the distributed file system is effectively improved, system architecture required by calling is simplified, any third-party mechanism/ system does not need to be depended, and the stability and the evolvability of the method and the device for calling the distributed file system are improved.

Description

technical field [0001] The present application relates to the field of computer technology, specifically to the field of Internet technology, and in particular to a method and device for invoking a distributed file system. Background technique [0002] A distributed file system running on a distributed file system needs to solve all the high-availability-related issues, such as master selection, master role notification, fence (Fence) mechanism, data flow replication and persistence, and data flow compaction (Compaction) and other requirements. [0003] The current distributed file system usually adopts the following two high-availability design schemes: [0004] 1. Combined solution: For example, choose a distributed service framework (ZooKeeper) solution, a heartbeat-based solution, or other self-designed solutions; Fence uses a third-party system or a self-designed solution; data flow replication uses self-designed replication program etc. [0005] 2. The overall solut...

AI Technical Summary

Problems solved by technology

[0006] However, the two high-availability solutions mentioned above, among them, the combined solution often relies on third parties, or even relies on multiple third parties at the same time. There are various self-designed solutions in the system, the overall structure of the system is complex, the stability risk is high, and It is difficult to evolve; in the overall solution, the distributed file system based on the Raft protocol is very slow to build a copy, and it is difficult to implement a fast load balancing mechanism, especially it is difficult to split and merge data streams on a global scale. file system collaboration

For example, a database (HBase) suitable for unstructured data storage, if the data flow uses the Raft protocol, then if it is still based on the distributed file system, it will cause a lot of additional traffic overhead, if it is no longer based on the distributed file system, then Will lose the stateless nature of HBase nodes

Images