This is a very common question we get. What’s going on behind the scenes when a node goes into warning status, or when it’s actually down? Well, it all starts with a simple ping. Orion pings its nodes at regular intervals to determine the status of those nodes. When Orion gets a response, everything is cool and the node is green.
What happens when Orion doesn’t get a response? When that happens, we can’t immediately assume the node is down, as there are a number of other reasons why the node may not be responding other than it being down. In other words, we want to be really sure the node is down before we mark it as such. How do we do this?
First, on the initial missed ping we are going to put the node into warning status by marking it yellow. When the node goes into a warning status, Orion initiates what we call a fast polling cycle where it pings the node every 10 seconds to try and determine if it’s really down. If Orion doesn’t receive any responses during this cycle, it determines the node is officially down and marks it red.
There is one other piece to this, which is how Orion calculates a node’s availability over time. What’s the difference between status and availability? To keep it very simple, a node’s status is an indicator of what the node is doing now; a node’s availability is a measurement of that node’s status over time. Orion calculates availability in two possible ways: either by historical node status, or by packet loss percentage over time. We give you the option of specifying which calculation to use. This setting is under Admin>>Settings>>Polling Settings (screenshot below).
I won’t go into a great deal of detail about these two settings, as you can read about them in the Administration Guide here; however, there are a couple of worthwhile points to mention about them. First, Node Status is the simpler calculation. It’s very straightforward; it basically looks at up/down status over time and calculates availability based on that. This leads me to my second point: use the Node Status setting unless you specifically need to know node availability based on packet loss. The Percent Packet Loss calculation is more complicated; don’t bother with it unless you specifically know you need it.