Some volumes, such as those that are used for database replay logs, are written sequentially and do not benefit from delayed cleaning of the DRL bits. For these volumes, sequential DRL can be used to limit the number of dirty regions. This allows for faster recovery should a crash occur. However, if applied to volumes that are written to randomly, sequential DRL can be a performance bottleneck as it limits the number of parallel writes that can be carried out.
The maximum number of dirty regions allowed for sequential DRL is controlled by a tunable as detailed in the description of
voldrl_max_seq_dirty in Tunable parameters.
See "Adding traditional DRL logging to a mirrored volume" on page 291.
See "Preparing a volume for DRL and instant snapshots" on page 285.