LRU values for flushing a buffer pool between checkpoints

The LRU values for flushing a buffer pool between checkpoints are not critical for checkpoint performance. The lru_max_dirty and lru_min_dirty values, which are set in the BUFFERPOOL configuration parameter, are usually necessary only for maintaining enough clean pages for page replacement. Start by setting lru_min_dirty to 70 and lru_max_dirty to 80.

If transactions are blocked during a checkpoint, the database server subsequently attempts to increase checkpoint frequency to eliminate the transaction being blocked. When the server searches for a free page to perform page replacement and a foreground write occurs, the server subsequently automatically increases the LRU flushing frequency to prevent this event from occurring again. When the database server completes page replacement and finds a frequently accessed page, the server automatically increases LRU flushing. Any automatic adjustments to LRU flushing do not persist to the onconfig file.

For more information about monitoring and tuning checkpoint parameters and information about LRU tuning and adjustments, see the HCL OneDB™ Performance Guide.