Browse Prior Art Database

Centralized management and simple failover method of filesystem resources

IP.com Disclosure Number: IPCOM000010448D
Original Publication Date: 2002-Dec-03
Included in the Prior Art Database: 2002-Dec-03
Document File: 1 page(s) / 8K

Publishing Venue

IBM

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 50% of the total text.

Page 1 of 1

Centralized management and simple failover method of filesystem resources

    Disclosed is a filesystem structure and a filesystem daemon program that enables it easy to construct a multi-node failover system. The only requirement for the nodes that will be part of the failover system is to install the filesystem daemon program on them. When the filesystem daemon detects the down of a certain node, it automatically takes over the failed node by executing programs in the super block of the filesystem. No configurations or no other special clustering program for each node is required.

The cluster failover system that is assumed here is the N-1 cluster system, in which 1 node is running as an active node and the other N nodes are running as standby node. And all of the nodes are attaching to the shared disks, but have non-shared filesystem on them. In case of using normal non-shared filesystem, in order to take over the node, the standby nodes must have the facility to know the down of the active node, the facility to make sure the failed node is not accessing the filesystem any more and the facility to restart the programs that have been running on the failed node. And also they must contain all the same programs and config files on the stanby node. It is very painful work to construct such a failover system. The filesystem disclosed here introduces the technique to simplify these managements.

This filesystem has a super block that consists of mainly the following three parts:
1. Application heartbeats area. This area is used for application heartbeats. Each application th...