prometheus API export data too large

[warrenincn]

apisix version: master
environment：K8S
scenario：Use apisix_bandwidth metrics to accomplish upstream's auto scale. Upstream's nodes use K8S pod ip. The pod ip changes a lot, lead to the prometheus export data too large(one route enable prometheus plugin, run 2 weeks, the export data now is 13839lines/1.7MB).
Relevant code：

apisix/apisix/plugins/prometheus/exporter.lua

Line 229 in 44d20c3

gen_arr("egress", route_id, service_id, consumer_name, balancer_ip))

In our scenario, the label: node(balancer_ip) is useless.
Any idea to settle this?

[tzssangglass]

ref: #4273

we are talking about that add some configuration items of the Prometheus plugin to support selecting what data to collect

[warrenincn]

How about put some label's value into fixed mock value according to route prometheus plugin's configuration? Some thing like below:

{
    "prefer_name":true,
    "access":[
        {
            "metric":"server_picker",
            "mock_label":[

            ]
        }
    ],
    "log":[
        {
            "metric":"bandwidth",
            "mock_label":[
                "node"
            ]
        }
    ]
}

[tzssangglass]

I think it would be more direct to let the user specify metrics