Whenever I select a time range lower than 30 minutes the dashboard for Metricbeat System doesn't seem to work properly. The gauges don't fill up at all and the network stats don't measure anything at all as well as the disk usage percentage, but the graphs seem to work fine as well as the number of processes.
Metricbeat System Overview - Time Range 10 minutes - Refresh 5 Seconds
Metricbeat Host Overview - Time Range 1 minute - Refresh 5 Seconds
Metricbeat System Overview - Time Range 30 minutes - Refresh 5 Seconds
Metricbeat Host Overview - Time Range 30 minutes - Refresh 5 Seconds
I have checked with many time ranges and it seems that the time limit is around 13 minutes, but even then the data shows up every other auto-refresh interval of 5 seconds. I also checked in the discover section and all the Metricbeat info is seems to be captured and processed correctly.
I am running ELK 6.2.2 as docker services inside a single-node swarm and the info is beign sent from other swarm services, each running a copy of Metricbeat and Filebeat 6.2.2. All of this is running inside a virtual machine running Red Hat Enterprise Linux 7.