Elastic Cloud on Kubernetes version 1.9.0

edit

Elastic Cloud on Kubernetes version 1.9.0

edit

Breaking changes

edit

For Operator Lifecycle Manager (OLM) and OpenShift OperatorHub users that run with automatic upgrades enabled, it is advised to set set-default-security-context operator flag explicitly before next ECK release. Next version of ECK will introduce a change that could result in setting an invalid Pod Security Context to Elasticsearch Pods.

  • Inherit resource requirements in init containers #4963

All init containers will inherit resource requirements from the main containers by default.

  • Move to env vars based Fleet mode configuration #4889

ECK operator will configure Elastic Agent in Fleet mode using environment variables instead of a configuration file. This method allows you to easily override the default settings. If the override is configured with any other method, an additional step is required before the operator upgrade.

  1. Update your Agent resources such that environment variables are set for settings you’d like to override, for example:

    ...
    spec:
      deployment:
        podTemplate:
          spec:
            containers:
            - name: agent
              env:
              - name: KIBANA_FLEET_HOST
                value: "https://kibana.example.com:443"
    ...
  2. Upgrade ECK to version 1.9.x.
  3. Remove any modification and references to fleet-setup.yml. As ECK 1.9.x uses environment variables to set the default config, the environment variables you’ve set in step 1.a. override the default values.

New features

edit
  • Allow users to suspend Elasticsearch Pods for debugging purposes #4946 (issue: #4546)

Enhancements

edit
  • Stop using deprecated xpack.monitoring.* settings #5136 (issue: #5083)
  • Stop using Enterprise Search deprecated setting in 7.17+ #5085 (issue: #5082)
  • Add logging for call to ES #4958 (issue: #4935)
  • Shorten the reconciliation loop duration if Elasticsearch is down #4938 (issues: #2939, #3496)
  • Ignore ClusterIPs, and IPFamilyPolicy in services, and timestamp in license during reconciliation #4929
  • Add operator license key check #4925
  • Skip remote clusters reconciliation earlier if remote clusters are not configured #4924
  • Refactor Elasticsearch HTTP error handling #4922 (issue: #4918)
  • Verify supported Elasticsearch distribution during license reconciliation #4920
  • Remove permissions not used by the operator #4823

Bug fixes

edit
  • Add kibana.host setting only to Enterprise Search 7.15+ #5122
  • Avoid updating the association status when no association #4986 (issue: #4985)
  • Forcibly recreate Beat keystore #4942 (issues: #4527, #4926)
  • Init. fs container: make cp more resilient #4888 (issue: #4877)
  • Allow the upgrade of a single node cluster with yellow status #4787 (issue: #4625)

Misc

edit
  • Update module github.com/go-test/deep to v1.0.8 #4948
  • Update golang Docker tag to v1.17.2 #4934
  • Update module github.com/spf13/viper to v1.9.0 #4847