To start with, there are four levels of severity for every attribute ( CPU, Disk space, JVM size etc are known as attributes ) except availability – Unknown, Critical, Warning and Clear. Availability has only two states – Up and Down.
By default any attribute will be in unknown state and once you associate a threshold it will become either critical / warning / clear depending upon the threshold. Also note that by default actions ( notifications like email, sms etc ) will be executed only when there is a change in level of severity ( ie., actions will be executed only when some attribute changes from clear to critical level. This action will not be executed again when that attribute remains in critical state ).
The Health attribute of a monitor or a monitor group will depend upon all the other attributes/monitor’s severity levels present under it by default. Only when all the attributes/monitors are in clear state, health will be in clear state. If any attributes becomes critical or if the availability of the monitor goes down, the health will automatically become critical.
Pros and Cons of configuring alert in a Monitor Group level
Lets say that you have configured availability alert for a monitor group. So when any one monitor in this group goes down, you will get an alert and also configuring alert in group level is easy work.
The problem in this approach is that when the first monitor goes down, the availability of the group will also become down and you will get email notification. If another monitor also goes down after some 5 minutes ( Monitor 1 is still down and availability of group is also already down ) you will not get any email notifications as there is no change in severity of the availability of monitor group.
The best approach to this problem is to configure availability alert for each monitor individually.
Pros and Cons of configuring alert for Health of a monitor
When you configure alert for health of a monitor, you will be alerted if any attribute under that monitor becomes critical. But this approach also has the same problem as configuring alert for monitor group availability. You will be notified when first attribute becomes critical or warning however you will not be notified if another attribute also becomes critical in the next poll ( when the first attribute is still critical ).
The best approach to this problem is again to configure alerts for each required attribute individually.
– Arun
Thanks, very interesting.