Description of problem:

In the "node-exporters" pods are full of messages where the collector is not able to read the InfiniBand counters:

2025-09-30T17:07:12.432674275+05:30 ts=2025-09-30T11:37:12.432Z caller=collector.go:169 level=error msg="collector failed" name=infiniband duration_seconds=0.006340497 err="error obtaining InfiniBand class info: failed to read file \"/host/sys/class/infiniband/qedr0/ports/1/counters/VL15_dropped\": invalid argument"

This error is already fixed in upstream in https://github.com/prometheus/node_exporter/issues/3265.

Version-Release number of selected component (if applicable):

Openshift 4.18.22

How reproducible:

Always

Steps to Reproduce:

1. Deploy Openshift running on top of a hardware with Infiniband
2. Review the "node-exporter" logs

Actual results:

1. Not counter metrics for Infiniband
2. for each node-exporter a high number of logs produced with the error exist. In ~40 minutes around 358 per "node-exporter" pod

   $ oc logs node-exporter-c8jxr  -c node-exporter -n openshift-monitoring|head -1
   2025-09-30T16:22:57.923447493+05:30 ts=2025-09-30T10:52:57+00:00 num_cpus=96 gomaxprocs=4
   
   $ oc logs node-exporter-c8jxr  -c node-exporter -n openshift-monitoring |tail -1
   2025-09-30T17:07:42.429386262+05:30 ts=2025-09-30T11:37:42.429Z caller=collector.go:169 level=error msg="collector failed" name=infiniband duration_seconds=0.000256877 err="error obtaining InfiniBand class info: failed to read file \"/host/sys/class/infiniband/qedr0/ports/1/counters/VL15_dropped\": invalid argument"
   
   $ oc logs node-exporter-c8jxr  -c node-exporter -n openshift-monitoring|grep -c "error obtaining InfiniBand"
   358
   
   $ for pod in $(oc get pods -l app.kubernetes.io/name=node-exporter -o name); do oc logs $pod -c node-exporter  ; done|grep -c "error obtaining InfiniBand"
   2148
   

Expected results:

1. Infiniband counter are collected as metrics
2. Not noisy errors in the "node-exporter" pods

Additional info: