Please define Group availability and Group Member availability

jpflanz

I am under the understanding that Group availability and Group member availability counts unmanaged time as downtime, thereby lowering true availablity.

I can make no sense of either measurement against any of the group of nodes  I have built

mesverrum

So a common assumption I run into is that availability is based on time, but orion doesn't have much of a sense of time, it doesn't keep an internal clock like that.  It just has datapoints for polled data, and the availability % is determined as a percent of the number of times we polled the object.

So in a normal day with a 2 minute polling intervals you have ~720 availability datapoints in the database that could be up or down.

If the node was down for 10 polls during that day then you have 98.61111% up time.

Now if you bring unmanaging into the mix, lets say you unmanaged the device for 1 hour, but it still was down for 10 polls during the 23 hours of the day that the node was managed.

Now we have only 690 availability datapoints to work with, and 10 of those were down, so we get 98.5507% up time.

Further impacting the numbers can be things like nodes with dependencies that become unreachable, because unreachable nodes are effectively unmanaged until their parent comes back up.  When a node goes into warning it triggers the fast polling which changes to a 10 second polling interval, or manually forced polls.

I've also seen cases where the numbers looked weird until we found out that the polling engine was overloaded or have some other problem that was making it so our completion rate was bad.

Essentially, what you need to do to verify the numbers is to generate a report of every availability datapoint during the interval, and do the math of [ count of up polls ] / [ count of total polls ]
Things get a little more complicated if you are going further back than 2 weeks because you would also need to factors weighting of the historical metrics into the calculation.