Werk #14613: Fix "Limit for upper bound dynamic levels" for predictive levels

You are affected by this change if you use the option "Limit for upper bound dynamic levels" for "Predictive Levels" for "CPU load".

If the predictive levels predict a very low value, it's easy to overshoot those levels. Therefore the "Limit for upper bound dynamic levels" can define warn and crit levels which are used when the predictive values are below this defined warn and crit levels.

The "CPU load" has a special mechanism that considers the number of CPUs, otherwise you would have to write special rules for servers with different number of CPU cores, as the load value is dependent on the number of cores. A high load for a 4 CPU computer may be a very low load for a 32 CPU computer.

That's why checkmks' load values are divided by the number of CPUs to be able to define one rule for all your computers independent of the number of CPUs.

The option "Limit for upper bound dynamic levels" did miss this adaption to the number of CPUs.

This has been fixed and now the "Limit for upper bound dynamic levels" is internally multiplied by the number of CPUs.

If you use the argument "Limit for upper bound dynamic levels" for "Predictive Levels" for "CPU load" and want to keep the current behaviour, you have to create different rules depending on the CPU count of your systems and divide the level entered by the number of CPUs. This is, of course not practical. We consider the old behaviour as not correct.

To the list of all Werks