Dynatrace Oneagent Operator Versions Save

Bug fixes

• Marked for Termination events are not a point in time instead of a time range of a few minutes (#229)
• Fixed error message when OneAgent has been already removed from the cache but the node was still there (#232)

Other changes

• Added environment variable 'RELATED_IMAGE_DYNATRACE_ONEAGENT' as preparation for RedHat marketplace release (#228)
• Fixed some problems with the current Travis CI build (#230)

Upgrading

The Operator can be upgraded from 0.7.0 with,

# Kubernetes
$ kubectl apply -f https://github.com/Dynatrace/dynatrace-oneagent-operator/releases/download/v0.7.1/kubernetes.yaml

# Openshift
$ oc apply -f https://github.com/Dynatrace/dynatrace-oneagent-operator/releases/download/v0.7.1/openshift.yaml

Improvements

• Added a setting to configure a proxy via the CR (#207)
• Added a setting to add custom CA certificates via the CR. Requires OneAgent image 1.39.1000 or newer (#208)
• No longer change the OneAgent .spec section to set defaults, easing integration with GitOps tools (#206)
• Show operator phase in the status.phase field of the OneAgent object (#197)
• Added proper error handling for Dynatrace API quota limit (#216)

Bug fixes

• Proxy environment variables (e.g., http_proxy, etc.) can be ignored on Operator container when skipCertCheck is true (#204)
• Istio objects don't have an owner object, so wouldn't get removed if the OneAgent object is deleted (#217)
• Handle sporadic (and benign) race conditions where the error below would appear (#194),

  Operation cannot be fulfilled on oneagents.dynatrace.com \"oneagent\": the object has been modified; please apply your changes to the latest version and try again
  

Breaking changes

• Operator images will be now published on Docker Hub (docker.io/dynatrace/dynatrace-oneagent-operator) instead of Quay.
• This version drops support for Kubernetes 1.11, 1.12, and 1.13; support for OpenShift 3.11, and 4.1, based on Kubernetes 1.11, and 1.13 respectively, is kept (#219, #220)
  • And while OpenShift 3.11 is supported, minimum version required is 3.11.188 unless a workaround is applied to the openshift.yaml manifest, see note on README.

Other changes

• Work in progress to add support for ARM64 environments (#201, #203):
  • Development snapshot images are now being published to Docker Hub
  • Build ARM64 images for the Operator (#201, #211)
  • Migrate CI/CD workflow from CircleCI to TravisCI
• Update to Operator SDK 0.15.1 (#200)
• Initial work to ease release automation (#198)
• Added automatic creation of CSV files for OLM (#210)
• Refactor of Marked for Termination feature, events will be sent only for deleted Nodes (#189, #196, #213, #214, #223)
• Support deprecation of beta.kubernetes.io/arch and beta.kubernetes.io/os labels (#199)
• Use v1 instead of v1beta1 for rbac.authorization.k8s.io objects (#215)

Upgrading

The Operator can be upgraded from 0.6.0 with,

# Kubernetes
$ kubectl apply -f https://github.com/Dynatrace/dynatrace-oneagent-operator/releases/download/v0.7.0/kubernetes.yaml

# Openshift
$ oc apply -f https://github.com/Dynatrace/dynatrace-oneagent-operator/releases/download/v0.7.0/openshift.yaml

Improvements

Additional fields have been added to the OneAgent CRD:

• Allow custom DNS Policy for OneAgent pods (#162)
• The service account for pods can now be customized (#182, #187)
• Custom labels can be added to pods (#183)

A schema has also been added to the OneAgent CRD:

• Add OpenAPI V3 Schema to CRD objects (#171)
• Validate tokens for OneAgent and show results as conditions on OneAgent status section (#188)

Other changes

• Operator Docker images have been merged, and are now the base image is UBI ([#179]
• Improve error logging from Dynatrace API requests on Operator (#185)
• Operator log entries now use ISO-8601 timestamps (e.g., "2019-10-30T12:59:43.717+0100") (#159)
• Most operations now use HTTP Header for authentication with Dynatrace API (#167) (https://github.com/Dynatrace/dynatrace-oneagent-operator/pull/179))
• Update to nested OLM bundle structure (#163)
• Code style improvements (#158, #175)
• Update to Operator SDK 0.12.0 and Go modules (#157, #172)
• Using istio.io/client-go to manage Istio objects (#174)
• Add OLM manifests for v0.6.0 (#193)

Breaking changes

• New OneAgent objects will be validated through the added schema by Kubernetes, so some issues may appear now, e.g., having an integer as tokens - it needs to be a YAML string now.
• From now on, we'll publish the installation manifests, kubernetes.yaml and openshift.yaml, as release attachments rather than having them on the repository code base. For this specific release however, we'll keep both.

Upgrading

The Operator can be upgraded from 0.5.x with,

# Kubernetes
$ kubectl apply -f https://github.com/Dynatrace/dynatrace-oneagent-operator/releases/download/v0.6.0/kubernetes.yaml

# Openshift
$ oc apply -f https://github.com/Dynatrace/dynatrace-oneagent-operator/releases/download/v0.6.0/openshift.yaml

Service update to get the latest version of the base image for RedHat Container Catalogue where some vulnerabilities have been fixed.

The Dynatrace OneAgent Operator now allows to customize the service account used for creating the OneAgent pods in preparation to release on GKE marketplace.

The Dynatrace OneAgent Operator v0.5.2 fixes a bug introduced in v0.5.1 where the Operator chooses the wrong Secret object for credentials.

The Dynatrace OneAgent Operator v0.5.1 includes a fix for Operator pod crashes when deploying the Operator on a cluster with unschedulable nodes not monitored by the OneAgent.

The Dynatrace OneAgent Operator v0.5.0 includes:

• Better detection and handling of node scaling events.
• Improved documentation for installation of the Operator in OCP 3.11 environments.

To upgrade from v0.4.0 you can run:

# Kubernetes
$ kubectl apply -f https://raw.githubusercontent.com/Dynatrace/dynatrace-oneagent-operator/v0.5.0/deploy/kubernetes.yaml

# OpenShift
$ oc apply -f https://raw.githubusercontent.com/Dynatrace/dynatrace-oneagent-operator/v0.5.0/deploy/openshift.yaml

This release version includes a bugfix -

Handles nil pointer panic in oneagent controller when reconciling istio objects - caused by no virtualservice created for IP-based communication hosts.

This release version includes bugfixes -

• Error handling when new CRDs are created after deploying operator (relevant in the case of istio-automanagement)
• Prevents operator from going into an infinite reconcile loop when no oneagent pods are deployed
• Creating appropiate Istio objects for HTTP and IP communication endpoints

Known Limitation

The enableIstio feature requires to restart the operator if Istio was deployed after deployment of the operator.

Background: This happens because the cache maintained by controller-runtime's Kubernetes Client is not dynamic. The bug for same is reported here https://github.com/kubernetes-sigs/controller-runtime/issues/321 and the fix for same is currently a work in progress https://github.com/kubernetes-sigs/controller-runtime/pull/554 .