I will try to explain a problem, before writing some suggestion.
In my example I have huge Prometheus output ~2000 metrics from kubernetes cluster.
Since we are talking about kubernetes, basically creating simple items is absolutely not an option. Instead Low Level Discovery should be used.

Currently I want to discover and create prototypes on this part:

kube_node_status_allocatable{node="node1",resource="memory",unit="byte"} 1.504894976e+09
kube_node_status_allocatable{node="node1",resource="pods",unit="integer"} 110
kube_node_status_allocatable{node="node1",resource="cpu",unit="core"} 1
kube_node_status_allocatable{node="node1",resource="ephemeral_storage",unit="byte"} 9.975103472e+09
kube_node_status_allocatable{node="node1",resource="hugepages_2Mi",unit="byte"} 0
kube_node_status_allocatable{node="node2",resource="memory",unit="byte"} 1.504894976e+09
kube_node_status_allocatable{node="node2",resource="pods",unit="integer"} 110
kube_node_status_allocatable{node="node2",resource="cpu",unit="core"} 1
kube_node_status_allocatable{node="node2",resource="ephemeral_storage",unit="byte"} 9.975103472e+09
kube_node_status_allocatable{node="node2",resource="hugepages_2Mi",unit="byte"} 0
kube_node_status_allocatable{node="master",resource="pods",unit="integer"} 110
kube_node_status_allocatable{node="master",resource="cpu",unit="core"} 2
kube_node_status_allocatable{node="master",resource="ephemeral_storage",unit="byte"} 1.9317915617e+10
kube_node_status_allocatable{node="master",resource="hugepages_2Mi",unit="byte"} 0
kube_node_status_allocatable{node="master",resource="memory",unit="byte"} 3.868737536e+09

I am interested only in memory, cpu and storage resources.
Since metric name is the same, I would want to manage all this within one single LLD.

I create a LLD rule with Prometheus to JSON

kube_node_status_allocatable{node=~".*",resource=~".*",unit=~".*"}

And filter

{#RESOURCE} matches memory|cpu|ephemeral_storage

Note, that I cannot use one single Item prototype because of multiple reasons

• Units of item itself are different in all 3 available resources
• Type of information is different (As example memory is float and cpu is integer)
• Trigger thresholds should be different for each resource
• etc

What I tried to do is:
create a prototype for memory and add prometheus pattern preprocessing, where i statically define that resource should be memory and units=byes for all discovered nodes.

{#METRIC}{node="{#NODE}",resource="memory",unit="byte"}

Obviously it doesn't work as expected. I still get items created for all discovered resources. Simply all of them have static resource and unit label in preprocessing steps.

From this result, I don't see any other way to achieve my result as only creating separate LLD rules for each resource.
In this kube_node_status_allocatable example I would have 3 LLD rules. But I have more then 1000 metrics available. Creating and maintaining all those LLDs seems like a nightmare.