30s css_misscount continues to frustrate customers when we have fabric recovery events

For the longest time we at Red Hat have been asking customer to change css_misscount to at least 65s to cater for fabric events and recovery. Fabric events and multipath recovery easily exceed the 27s watchdog so called short disk timeout (css_misscount-3) and cause false evictions.

Many customer now do this but every time we get push-back from Oracle frontline saying it is not allowed to be changed. When we however get an Oracle back-line resource engaged they agree that certain fabric events will need tuning.

We really need you to change your documentation and show that we have scenarios where one has to tune the css_misscount to higher timeout values and not always have a frontline resource say not allowed to be changed.