dashboards/vmcluster: vmauth reported as vmselect and vminsert

[rtm0]

For some reason VictoriaMetrics - cluster dashboard treats vmauth as both vminsert and vmselect instances. This manifests in all charts that include vmselect and/or vminsert. Some examples where it is noticed right away:

1. Stats section at the very top. The version and uptime charts show additional vmselects and vminserts:

2. Drilldown section at the very bottom. These additional vmselects and vminserts are included into the CPU Usage chart. This is where you can see that it is actually a vmauth instance - both vmselect and vminsert have the same IP and 8427 port.

To Reproduce use our sandbox dashboard and choose lts deployment.

---

If vmauth consumes little resources, then its presence may be not noticed on other charts. But if it has significant resource utilization this will show as if vmselect or vminsert consume more resources that they actually do.

Additionally, even if vmauth consumes little resources, but has some utilization spikes (i.e. its max cpu is 0.5, but it suddenly uses 0.45 cpu) it will significantly distort vmselect and vminsert utilization charts.

An example from my personal cluster:

Here is the Resource Usage / CPU (All) chart. Notice the vmselect spikes (yellow line). The vminsert has exactly same spikes (the green line, it is hidden behind yellow line because it is exactly the same):

If you look at the same chart but unaggregated, we will see that vmselect and vminsert actually have no spikes. Instead these are vmauth spikes.

Because this chart does not use absolute numbers and instead uses the relative ones (current utilization-to-max), vmauth utilization wins (because the query also has max function).

[f41gh7]

Looks like it's related to the requestsLoadBalancer feature and VMServiceScrape objects generated by operator.

[AndrewChubatiuk]

bug was fixed in 0.67.0