evaluating Rancher for use in large-scale iot-edge solution where most of the edge locations will have satellite connectivity. Using nethogs and ifstat i have measured that between k3s and Rancher the bandwith usage is about 2.6 KiloBytes /s (KB/s).
-i use two hyper-v instances with linux, one with Rancher in docker and one with k3s.
-The k3s cluster doesn’t have any load/apps except for the Rancher agents.
-Measuring network usage directly on the k3s host machine.
-No monitoring enabled in the Rancher portal, ie. forwarding of prometheus stats not configured.
-rancher version 2.5.8
-If i kill the fleet pod the bandwidth usage drops to about 2.4 KiloBytes /s (KB/s)
-if i delete the cattle cluster agent the bandwidht usage drops to zero.
So it seems to me that the majority of the bandwidth is used by the the cattle cluster agent somehow.
Is there a way to tune this ? Our customers are very cost sensitive when it comes to use of satellite bandwidth, since at the moment this is very expensive. It would be ok if we don’t have continous RT updated overviews in Rancher, but we probably cannot use it if we cannot bring down the continous bandwidth drag …