When attempting to rollforward from a checkpoint, Im getting memory errors – how can I reduce the memory requirements?
Rollforwarddb allocates memory to create its own local caches from 2k through to 64k whether or not they are being used in your installation. The default size for the 2k cache is 256 buffers and for the 4k, 8k, 16k 32k and 64k caches is 200 buffers. However if a value of 0 is specified for any of the caches this results in a cache size of 256 buffers being created. To reduce the memory requirements of the 64k cache for example, you need to specify the flag -dmf_cache_size_64k=1 which will set the 64k cache to 1 buffer only. Similarly, -dmf_cache_size_32k=0 and so on can also be used to reduce the size of the other caches that won’t be needed. The DMFJSP process uses its own cache independent of the DBMS one, to ensure that it has a cache even for tables of page sizes that may have been used and then removed while re-applying the journals. If you know there are no pages of a particular size in the journals to be recovered from, these parameters are safe to use.
