Data recovery and high availability

High availability solutions using multi-instance queue managers must include a mechanism to recover data after a storage failure.

A multi-instance queue manager increases the availability of queue manager processes, but not the availability of other components, such as the file system, that the queue manager uses to store messages, and other information.

One way to make data highly available is to use networked resilient data storage. You can either build your own solution using a networked file system and resilient data storage, or you can buy an integrated solution. If you want to combine resilience with disaster recovery, then asynchronous disk replication, which permits disk replication over tens, or hundreds of kilometers, is available.

You can configure the way different IBM® MQ directories are mapped to storage media, to make the best use of the media. For multi-instance queue managers there is an important distinction between two types of IBM MQ directories and files.
Directories that must be shared between the instances of a queue manager.
The information that must be shared between different instances of a queue manager is in two directories: the qmgrs and logs directories. The directories must be on a shared networked file system. You are advised to use a storage media that provides continuous high availability and excellent performance because the data is constantly changing as messages are created and deleted.
Directories and files that do not have to be shared between instances of a queue manager.
Some other directories do not have to be shared between different instances of a queue manager, and are quickly restored by means other than using a mirrored file system.
  • IBM MQ executable files and the tools directory. Replace by reinstalling or by backing up and restoring from a backed up file archive.
  • Configuration information that is modified for the installation as a whole. The configuration information is either managed by IBM MQ, such as the mqs.ini file on Windows, UNIX and Linux® systems, or part of your own configuration management such as MQSC configuration scripts. Back up and restore using a file archive.
  • Installation-wide output such as traces, error logs and FFDC files. The files are stored in the errors and trace subdirectories in the default data directory. The default data directory on UNIX and Linux systems is /var/mqm. On Windows the default data directory is the IBM MQ installation directory.
You can also use a backup queue manager to take regular media backups of a multi-instance queue manager using linear logging. A backup queue manager does not provide recovery that is as fast as from a mirrored file system, and it does not recover changes since the last backup. The backup queue manager mechanism is more appropriate for use in off-site disaster recovery scenarios than recovering a queue manager after a localized storage failure.