Uploaded image for project: 'OpenShift Request For Enhancement'
  1. OpenShift Request For Enhancement
  2. RFE-4708

Ability to customize also the diskstats collector of node-exporter

    XMLWordPrintable

Details

    • Feature Request
    • Resolution: Unresolved
    • Normal
    • None
    • None
    • Monitoring
    • None
    • False
    • None
    • False
    • Not Selected
    • 0
    • 0% 0%

    Description

      Following with the customization we have already and that are mentioned in:

      https://issues.redhat.com/browse/MON-2814
      https://issues.redhat.com/browse/MON-2848

      IHAC who would need to enable/disable or parametrize diskstats collector. Here it's the justification:
      =====================================
      They are using a lot (800+) of fiberChannel connected storage devices in our 60+ node OCP4 cluster. These devices are connected to each of the 60+ workers via multipath and create 1600+ device entries under /dev/...

      The node-exporter of prometheus creates a high load for collecting the disc-stats of these devices (every 15 secs), which impacts performance of the whole system.
      =====================================

      we observer node-exporter is starting to collect diskstats for every "local" device every 15 secs and sometimes it has not enough time to finish before the next run starts.

      This is quite related to:

      https://issues.redhat.com/browse/OCPBUGS-19495
      kubelet and node_exporter consumption degrading performance from OCP worker nodes

      Attachments

        Issue Links

          Activity

            People

              rh-ee-rfloren Roger Florén
              rhn-support-gparente German Parente
              Votes:
              1 Vote for this issue
              Watchers:
              5 Start watching this issue

              Dates

                Created:
                Updated: