With mono data storage, data cannot be read from the cache medium following a cache memory error and any write data in the cache will be lost unless it has been written to disk.
If a memory error occurs when accessing a main memory cache segment, the data access operation in question will, if possible, be redirected to the relevant external data memory. An I/O error will be generated if the data access operation cannot be redirected, for example because the only up-to-date version of the data to be accessed is in the cache memory.
DAB indicates errors of this type by means of messages on the console. These messages are of the type “asynchronous question” and are output for each cache area for which an error was found.
DAB automatically starts to write cache data to disk for areas where an error was detected. The affected area will be released once the cache data has been successfully written to disk. Even if the area cannot be released, it remains locked against further caching of data.
Message indicating an access error:
NDB0043 THE FOLLOWING CACHE BUFFER CONTAINS DEFECTIVE BLOCKS: '(&00)'
No data will be lost when using cache areas with read caching as all data will be on both the disk and in the cache. Cache areas can be released using the command /STOP-DAB-CACHING.
When using cache areas with write or read/write caching, the cache may contain defective blocks for which there is no equivalent on disk. In this case, the data in the cache can no longer be read and the blocks are lost. The affected files can only be reconstructed if a data backup is available, and then only to the version saved in that backup. When caching entire volumes in write mode, the affected volume may need to be reinitialized. The affected data area can be found using the command /SHOW-DAB-CACHING for the relevant cache area. When caching using AREA=*FILE, files are flagged by the status PINNED DATA.