IBM Books

Managing Shared Disks

Preserving data integrity during application recovery

Rather than using hc, you can use fencing. Application recovery can be independent.

If a recoverable database application is running in a system and one of the nodes in that system fails, any data that had been locked by the failed application instance is returned to a consistent state. If the node that failed is still capable of issuing I/O requests to virtual shared disks, those requests must not be carried out. An application's recovery script can use the fencevsd command to prevent I/O operations from failed nodes to virtual shared disks in the system.

When the node that failed is active again, the application's recovery script can issue the unfencevsd command to permit it to issue I/O requests to virtual shared disks again.

You can check the fenced status of all the nodes in the system with the lsfencevsd command, which displays a map of all fenced virtual shared disks and the nodes that are fenced from them.

[ Top of Page | Previous Page | Next Page | Table of Contents | Index ]