Recovering CF structures

At least one queue manager in the queue-sharing group must be active to process a RECOVER CFSTRUCT command. CF structure recovery does not impact queue manager restart time, because recovery is performed by an already active queue manager.

The recovery process consists of two logical steps that are managed by the RECOVER CFSTRUCT command:

1. Locating and restoring the backup.

2. Merging all the logged updates to persistent messages that are held on the CF structure from the logs of all the queue managers in the queue-sharing group that have used the CF structure, and applying the changes to the backup.

The second step is likely to take much longer because a lot of log data might need to be read. We can reduce the time taken if you take frequent backups, or if you recover multiple CF structures at the same time, or both.

The queue manager performing the recovery locates the relevant backups on all the other queue managers' logs using the data in DB2 and the bootstrap data sets. The queue manager replays these backups in the correct time sequence across the queue sharing group, from just before the last backup through to the point of failure.

The time it takes to recover a CF structure depends on the amount of recovery log data that must be replayed, which in turn depends on the frequency of the backups. In the worst case, it takes as long to read a queue manager's log as it did to write it. So if, for example, you have a queue-sharing group containing six queue managers, an hour's worth of log activity could take six hours to replay. In general it takes less time than this, because reading can be done in bulk, and because the different queue manager's logs can be read in parallel. As a starting point, we recommend that you backup your CF structures every hour.

All queue managers can continue working with non-shared queues and queues in other CF structures while there is a failed CF structure. However, if the administration structure has also failed, all the queue managers in the queue-sharing group must be started before we can issue the RECOVER CFSTRUCT.

Backing up CF structures can require considerable log writing capacity, and can therefore impose a large load on the queue manager doing the backup. Choose a lightly-loaded queue manager for doing backups; for very busy systems, add an additional queue manager to the queue-sharing group and dedicate it exclusively for doing backups.