Conversation
|
Ugh, so sorry I missed this. Just getting round to catching up and working on the collector again. |
|
@tenortim you can go ahead and implement the changes within your code. It looks like some of the modifications I have made in my fork made irrelevant anyway. As for the label, it may be good to a config option for like "cluster_label_name" or something like that where the user could configure the desired label name that is used. In a perfect world, I think "cluster" makes the most sense in terms of the Isilon/PowerScale metrics. The only issue is it is frequently used with the replica label for deduping data from multiple Prometheus instances. |
First, I ran
go mod tidyto correct the influxdb1-client being shows as an indirect module instead of being shown as a direct module.Other than that, I made some changes to the prometheus code:
instancewith the cluster name. The main reason is that the labelclusteris frequently used when running prometheus in kubernetes to denote which kubernetes cluster prometheus is running in. Since this is used for deduplication of metrics, it overrides theclusterlabel in the metrics resulting in nothing other than the endpoint name and port to identify which isilon cluster is which. Note that theinstancelabel will come across asexported_instance. I also left theclusterlabel for backwards compatibility.