feat: add metric type 'buckets'

Author: b-abadie

README.md

@@ -437,6 +437,44 @@ A `mapping` metric looks like this:
 arista_show_ip_bgp_summary_peerState{hostname="198.51.100.1",name="router-1.example.com",peer="192.168.1.1",vrf="default"} 6.0
 ```
 
+#### Buckets
+
+```yaml
+  show ip route vrf all summary:
+    type: multiple
+    lookup_keys: vrfs
+    metrics:
+      - name: maskLen
+        type: buckets
+        bucket_name: prefix_length
+```
+
+Some data are spread out in a histogram kind of way. This is the case with the `maskLen` metric of `show ip route summary` 
+which show number of routes spread out per mask length:
+
+```json
+{
+    "vrfs": {
+        "default": {
+            "maskLen": {
+                "8": 2,
+                "31": 19,
+                "32": 25
+            }
+        }
+    }
+}
+```
+
+Instead of going through every possible metric from `- name: maskLen.0` up to `- name: masklen.32`, a metric of type `buckets`
+with a `bucket_name` will automatically produce metrics for every possible values represented as labels, like this :
+
+```
+arista_show_ip_route_vrf_all_summary_maskLen{hostname="198.51.100.1",name="router-1.example.com",prefix_length="8"} 2.0
+arista_show_ip_route_vrf_all_summary_maskLen{hostname="198.51.100.1",name="router-1.example.com",prefix_length="31"} 19.0
+arista_show_ip_route_vrf_all_summary_maskLen{hostname="198.51.100.1",name="router-1.example.com",prefix_length="32"} 25.0
+```
+
 ### Labels
 
 #### prom_name

api_commands.yaml

@@ -92,6 +92,9 @@ commands:
       - name: dynamicPolicy
       - name: internal
       - name: ospfCounts.ospfTotal
+      - name: maskLen
+        type: buckets
+        bucket_name: prefix_length
     labels:
       - name: vrfs
         prom_name: vrf

arista-eapi-exporter.py

@@ -363,6 +363,14 @@ def main():  # pylint: disable=missing-function-docstring
                     labelnames=host_labels + normalized_labels,
                     states=metric["enum"],
                 )
+            elif metric_type == "buckets":
+                exported_metrics[metric_name] = Gauge(
+                    metric_name,
+                    f"Arista EOS metric '{metric['name']}' under '{command}'",
+                    labelnames=host_labels
+                    + normalized_labels
+                    + [metric["bucket_name"]],
+                )
 
     # This will hold each set of label:value ("labelset") PREVIOUSLY KNOWN for each metric. At each poll cycle, it will be compared
     # with the retrieved label:value set, in order to remove the no-longer-valid ones.
@@ -575,10 +583,12 @@ def main():  # pylint: disable=missing-function-docstring
                             exported_metrics[metric_name].labels(
                                 **extracted_labels
                             ).set(metric_data)
+
                         elif metric_type == "enum":
                             exported_metrics[metric_name].labels(
                                 **extracted_labels
                             ).state(metric_data)
+
                         elif metric_type == "mapping":
                             mapped_value = metric["mapping"].get(metric_data)
                             if mapped_value is None:
@@ -594,6 +604,13 @@ def main():  # pylint: disable=missing-function-docstring
                                 **extracted_labels
                             ).set(mapped_value)
 
+                        elif metric_type == "buckets":
+                            for bucket_value, bucket_data in metric_data.items():
+                                bucket = {metric["bucket_name"]: bucket_value}
+                                exported_metrics[metric_name].labels(
+                                    **extracted_labels, **bucket
+                                ).set(bucket_data)
+
                         labelsets_current[metric_name].append(extracted_labels)
 
             logger.info("Finished polling %s", target["name"])