Uploaded image for project: 'OpenShift Monitoring'
  1. OpenShift Monitoring
  2. MON-3237

Tolerate scrape timestamp jitter to optimize metrics compression


    • Icon: Epic Epic
    • Resolution: Done
    • Icon: Major Major
    • openshift-4.15
    • None
    • None
    • scrapetime jitters
    • False
    • None
    • False
    • Not Selected
    • NEW
    • To Do
    • NEW
    • 0% To Do, 0% In Progress, 100% Done
    • 0

      Proposed title of this feature request

      Add scrape time jitter tolerations to prometheus

      What is the nature and description of the request?

      Change the configuration of the OpenShift Prometheus instances to tolerate scrape time jitters.

      Why does the customer need this? (List the business requirements)

      Prometheus chunk compression relies on scrape times being accurately aligned to the scrape interval. Due to the nature of delta of delta encoding, a small delay from the configured scrape interval can cause tsdb data to occupy significantly more space.

      We have observed a 50% difference in on disk tsdb storage for a replicated HA pair.

      The downside is a reduction in sample accuracy and potential impact to derivatives of the time series. Allowing a jitter toleration will trade off improved chunk compression for reduced accuracy of derived data like the running average of a time series.

      List any affected packages or components.


            jfajersk@redhat.com Jan Fajerski
            jfajersk@redhat.com Jan Fajerski
            Junqi Zhao Junqi Zhao
            0 Vote for this issue
            7 Start watching this issue