Right now the health of a group is based on a static selection of hosts. If one purpose of groups is to provide business service health of a cluster of servers, I shouldn't have to designate specific servers as health indicators.
If I have a server farm of 9 servers, and 2 servers are down, technically the business services provided by that farm are still functional but the farm requires someone's attention or action.
I should be able to set a percentage or fraction and use the same alert severities in the system:
Examples:
Percentages:
- If 20% of my servers are critical, the group is at attention.
- If 60% are critical, we are at trouble
- 90% is critical.
Or fractions:
- If 1/5 of my farm is critical, attention, 3/5th, trouble, 4/5ths critical.
You can also add some more logic based on each systems worse alarm severity:
- If 50% of the systems are in at "attention" have that role up.
- If 25% of the systems are at "critical", have that role up to the group.
Otherwise, I could have a group with 50 systems and when 1 of them has any alert severity, the group severity matches.
Regardless, please make it so we don't select static systems for group health.