heistp
commented
2 years ago Hi, I haven't run the smokeping probe with similar parameters (i.e. a large number of pings at a lower interval), but are you certain that this doesn't have to do with the way smokeping chooses its colors? There is a large gap between purple at 4/600 and pink at 300/600. One would expect it to choose pink only when there were at least 300/600 losses during one step time, but it may be choosing pink more often than we would expect. In my experience, smokeping seems to select more pessimistic colors for the longer timescales. See the picture below for an example of this using the standard ping probe without irtt. It's not a great example and yours is more extreme, but the event on Tuesday appears less severe in the "Last 30 Hours" graph than the "Last 10 Days" graph.
How does it look for you in the "Last 3 Hours" graph?
Also, since you're using a much larger number of pings than the default plot configuration expects, I wonder what would happen if you configure smokeping's loss_colors to add more levels of loss than just 0, 1, 2, 3, 4, 300(big gap!) and 599. Those are just some ideas, anyway... :)
Trying to setup the smokeping probe for irtt, but I'm running into some odd behavior. In the graph, it implies I'm losing about 50% of packets all the time. While my internet is indeed quite bad, it's not that bad. The confusing part is on the lower bit of the graph, it will say only like 1% loss at a period where the graph shows the 50% loss. ie. 300/600 packets dropped. I'm certain I've messed up some configuration here, but don't quite understand where I've gone wrong.
My configuration and an example of what I'm seeing are attached.
smokeping-irtt.txt