Physical log size

How one can know correct size of physical log? Sun solaris 5.7 IDS 7.31 UC2A TIA

"Shevkar, Mukund" wrote: > > How one can know correct size of physical log? > Sun solaris 5.7 > IDS 7.31 UC2A The physical log must hold enough pages under all but the most extreme load (and one could argue even then) that the physical log reaching 75% full, triggering an early checkpoint) will not occur before a scheduled checkpoint occurs to clear out the log. Making it too large wastes disk space but not too badly. You can use 'onstat -l -r 5' to monitor the physical log usage from checkpoint to checkpoint. If your checkpoints are happen ONLY on schedule (ie the time interval between the beginning of one checkpoint and the next is CKPTINTVL + the duration of the earlier checkpoint) then the physical log is at least large enough (check during data loads and other peak update activity periods). If not a quick way to calculate how big to make it is to divide CKPTINTVL by the actual gap between the premature checkpoints and multiply that by PHYSFILE and divide the result by 0.75. The result is your new PHYSFILE value. Now monitor this again to determine if it needs to be adjusted downward or even further upward. If you want to start from the other direction, ie estimate PHYSFILE based on estimated load, the calculation is in the Adminstrator's Guide. Art S. Kagel