A common mistake made when using Vdbench data validation.
By Henk Vandenbergh on Oct 22, 2008
When using Vdbench data validation, the content of a data block is validated only when this block is accessed for the second or third etc., time. This means that if you have a very large LUN or file, and your run has a relatively small elapsed time, a run can complete without any data block ever having been accessed more than once. This means then that the run appears successful, but no real data validation has ever been done. Now of course you can have undetected data integrity problems!
I therefore decided to make a change in Vdbench 5.00: I will count the amount of data block validations that have been done, and if at the end of a run that count still is zero, Vdbench aborts Vdbench with an explanation about what happened and suggestions on how to fix this, things like:
- Use larger block sizes
- Use longer elapsed= times.
- Use only a portion of the LUN, using:
- sd=sd1,….,range=(xx,yy) (Vdbench 5.00)
08:31:10.258 No read validations done during a Data Validation run.
08:31:10.258 This means very likely that your run was not long enough to
08:31:10.258 access the same data block twice.
08:31:10.258 There are several solutions to this:
08:31:10.259 - increase elapsed time.
08:31:10.259 - use larger xfersize.
08:31:10.259 - use only a subset of your lun by using the 'sd=...,size='
08:31:10.259 parameter or the 'wd=...,range=' parameter.
Normal completion message on logfile.html:
08:33:23.847 localhost-0: Total amount of blocks read and validated: 1271