Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-56251

[release-4.18] Prepare for alertmanager v1 no longer supported in Prometheus v3

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Done-Errata
    • Icon: Major Major
    • 4.18.0
    • 4.18.z
    • Monitoring
    • None
    • Quality / Stability / Reliability
    • False
    • Hide

      None

      Show
      None
    • 4
    • None
    • None
    • None
    • None
    • MON Sprint 269, MON Sprint 270, MON Sprint 271, MON Sprint 272
    • 4
    • In Progress
    • Bug Fix
    • Hide
      Previously, using Alertmanager apiVersion v1 in openshift-monitoring/cluster-monitoring-config or openshift-user-workload-monitoring/user-workload-monitoring-config caused OCP 4.19 to fail early (InvalidConfigXXX) as OCP 4.19 is using Prometheus V3 and Alertmanager apiVersions v1 is not supported in it. With this release, to prevent the InvalidConfigXXX error after upgrading to OCP 4.19, CMO sets upgradable=false if apiVersion: v1 is detected in the config maps ensuring clusters upgrade through OCP 4.18.x before moving to OCP 4.19
      Show
      Previously, using Alertmanager apiVersion v1 in openshift-monitoring/cluster-monitoring-config or openshift-user-workload-monitoring/user-workload-monitoring-config caused OCP 4.19 to fail early (InvalidConfigXXX) as OCP 4.19 is using Prometheus V3 and Alertmanager apiVersions v1 is not supported in it. With this release, to prevent the InvalidConfigXXX error after upgrading to OCP 4.19, CMO sets upgradable=false if apiVersion: v1 is detected in the config maps ensuring clusters upgrade through OCP 4.18.x before moving to OCP 4.19
    • None
    • None
    • None
    • None

      Prometheus v3 no longer supports alertmanager v1

      OCP 4.19 will fail early (InvalidConfigXXX) in case alertmanager apiVersion: v1 is still specified in one of the two configmaps, see https://github.com/openshift/cluster-monitoring-operator/pull/2544/files

      To avoid that InvalidConfigXXX from triggering just after the 4.19 upgrade, we can make users bump their alertmanagers and only use apiVersion: v2 in the configmaps.

      We can do that by making CMO set upgradable=false when it sees apiVersion: v1 in one of the configmaps and make sure clusters go through that 4.18.x before upgrading to 4.19.

      One can get inspired by the following PRs
      https://github.com/openshift/cluster-monitoring-operator/pull/2544/files
      https://github.com/openshift/cluster-monitoring-operator/pull/2494/files
      https://github.com/openshift/cincinnati-graph-data/pull/6265

              janantha@redhat.com Jayapriya Pai
              rh-ee-amrini Ayoub Mrini
              None
              Ayoub Mrini
              Junqi Zhao Junqi Zhao
              None
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

                Created:
                Updated:
                Resolved: