This section very briefly explains some aspects of IMS logging and how its configuration can affect the Log Capture. IBM publishes a great deal more on the subject and you may want to browse their documentation for more details.
The IMS Log Capture reads the IMS transaction logs looking for special type "99" records placed there by IMS for the explicit use by Capture agents.
The IMS online log datasets (OLDS) are made up of multiple physical datasets and are written in a wraparound fashion. When the OLDS is full the contents are archived to the System Log Dataset (SLDS). The physical datasets are manually added and removed. Database Recovery Control (DBRC) assists in the recovery of databases, maintenance of the OLDS and also cleans up SLDS when they are no longer needed. DBRC stores information required for recovery in RECON datasets
The goal of the Log Capture is to read data only from the OLDS for best performance. If capture has been down for a while or IMS data is being captured much faster than subscribing Engines, the capture may have to drop back to read SLDS.