0.7.2

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.7.2/manifests/install.yaml

Changes since v0.7.1

Rollouts

• Update RS if RS's annotations need to be changed #413

v0.7.1

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.7.1/manifests/install.yaml

Changes since v0.7.0

General

• Adding ca-certificates to docker image (#393)
• Add patchMergeKey and patchStrategy struct tags and comments (#386)
• Reduce log from error to warning (#394)

Experiments

• Allow ex to use availableAt and finishedAt as args (#400)
• Experiments passed duration succeed with running analysis (#392)
• Refactor Experiment handling of pod hashes (#385)

v0.7.0

v0.7.0

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.7.0/manifests/install.yaml

Important Notices

• Please upgrade to v0.6.x before upgrading to v0.7. Pre v0.6.0 has a different pausing logic, and v0.7.0 removes the depreciated PauseStartTime field. The v0.6.x versions have a migration script that is removed in v0.7.0.
• This release introduces an alpha implementation of Rollouts leveraging Istio resources for traffic shaping. Check out traffic management for more info.

Changes since v0.6.3

General

• Support instance ids for rollout controller segregation #342
• Remove PauseStartTime #349
• Vendor mockery utility #347
• Remove loud log message #333

Rollouts

General

• Add stableService field #337

Traffic Routing

• Initial Istio implementation #341
• Implement watch for Istio resources #354
• Add validation to istio virtual services #355

Kubectl Plugin

• Introduce 'kubectl argo rollouts terminate' command #297

Analysis

Genernal

• Allow controller to delay analysis #350
• Create one background analysis per revision #309
• Allow AnalysisRun to complete an experiment #345

Providers

• Wavefront metric provider #338
• Web metric provider #318
• Refactor common logic in providers to library #368
• Allow web provider to be parameterized #368

v0.6.3

v0.6.3

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.6.3/manifests/install.yaml

Changes since v0.6.2

Bug Fixes

• Fix premature scaledown (#365)
• Add namespace restriction to job informer (#362)
• Fix honoring autoPromotionSeconds (#360)
• Ensure podHash stays on stable-svc selector (#340)

v0.6.2

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.6.2/manifests/install.yaml

Bug Fixes

• omitted revisionHistoryLimit was not defaulting to 10 (#330)
• Fix panic if rollout cannot create a new RS (#328)
• Enable controller to handle panics with crashing (#328)

v0.6.1

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.6.1/manifests/install.yaml

Changes since v0.6.0

Bug Fixes

• Create one background analysis per revision (#309)
• Fix Infinite loop with PreviewReplicaCount set (#308)
• Fix a delete by zero in get command (#310)
• Set StableRS hash to current if replicaset does not actually exist (#320)
• Bluegreen: allow preview service/replica sets to be replaced and fix sg fault in syncReplicasOnly (#314)

v0.6.0

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.6.0/manifests/install.yaml

Important Notices

• The pause functionality was reworked in the v0.6 release. Previously, the .spec.paused field was used by the controller to pause rollouts. However, this was an issue for users who wanted to manually pause the rollout since the controller assumed it was the only entity that set the field. In v0.6, the controller will add a pause condition to the .status.pauseCondition to pause a controller instead of setting spec.paused. The pause condition has a start time and a reason explaining why it paused. This allows users to set the spec.paused field manually and let the controller respect that pause. The v0.6 controller has a migration function to convert pre v0.6 rollouts to the new pause condition. The migration function will be removed in a future release.
• In pre-v0.6 versions, the BlueGreen strategy would have the preview service point at no ReplicaSets if the new ReplicaSet was receiving traffic from the active service. V0.6 changes that behavior to make the preview service always point at the latest ReplicaSet

Changes since v0.5.0

General

• Update k8s library dependencies to v1.16 (##192)

Rollouts

Enhancements

• Add Rollout Context to reconciliation loop (##205)
• Refactor pausing (##211)
• Allow User pause (##216)
• Stop progress while paused (##193)
• Add pause condition migration (##229)
• Add abort functionality (##224)
• Rollout analysis plumbing (##183)
• Add AnalysisStep for Rollouts (##188)
• Add background analysis runs for rollouts (##196)
• Clean up old Background AnalysisRuns (##246)
• Clean up Experiments and AnalysisRuns (##197)
• Add initial Experiment Step (##165)
• Make specifying replicas/duration optional in the experiment step (##241)
• Terminate experiments from previous steps (##280)
• Add Analysis to RolloutExperimentStep (##238)
• Fix TimeOut check to consider experiment/analysis steps (##278)
• Pause a rollout upon inconclusive experiment (##256)
• Abort a rollout upon a failed experiment (##256)
• Add create AnalysisRun action in clusterrole (##231)

Bug Fixes

• Fix nil ptr for newRS (##233)
• Fix Rollout transformer config (##247)
• Always point preview service at the newRS (##217)
• Make active service required (##235)
• Reset ProgressDeadline on retry (##282)
• Ignore old running rs for RolloutCompleted status (##218)
• Remove scale down annot after scaling down (##187)
• Renames golang field names for blueGreen/canary to eliminate two API violations (##206)

Experiments

Check out the Experiment Docs for more information.

Enhancements

• Refactor experiments to use a context object (##208)
• Allow selectors to be overwritten when starting experiments (##249)
• Simplify experiment replicaset names (##274)
• Integrate Experiments with Analysis (##210)

Bug Fixes

• Fix experiment enqueue logic (##239)
• Annotate instead of label experiment names in replicasets (##262)
• Fix issue where a replicaset name collision could cause hot loop (##236)

Analysis

Check out the Analysis Docs for more information.

Enhancements

• AnalysisRun AnalysisTemplate Spec (##166)
• Initial analysis controller implementation (##168)
• Integrate analysis controller with provider interfaces (##171)
• Add metric knob for maxInconclusive (##181)
• Simplify provider interfaces to set error messages (##189)
• Implement ResumeAt logic (##232)
• Define explict args in AnalysisTemplates and simplify AnalysisRun spec (##283)
• Use a duration string instead of int to represent duration (##286)
• Truncate measurements when greater than default (10) (##191)
• Add counter for consecutiveError (##191)

Prometheus Provider

• Add initial provider and Prometheus implementation (##170)
• Rename prometheus.server to address to better reflect API client interface (##207)
• Treat NaN as inconclusive on Prometheus provider (##275)

Job Provider

• Implement job-based metric provider (##186)
• Job metric argument substitution. Simplify metric provider interface (##268)

Kayenta Provider

• Initialize check in for kayenta metric provider ##284

Kubectl Plugin

Check out the kubectl plugin docs for more information.

Enhancements

• Implement argo rollouts kubectl plugin (##195)
• Introduce kubectl argo rollout list rollouts command (##204)
• Introduce kubectl argo rollout list experiments command (##267)
• Introduce kubectl argo rollout set image command (##251)
• Introduce kubectl argo rollout get command (##230)
• Introduce kubectl argo rollout promote command (##277)
• Add ability to kubectl argo rollouts set image *=myrepo/myimage (##290)
• Add get/retry experiment commands. Support experiment retries (##263)
• Show running jobs as part of analysis runs (##278)
• Surface experiment images to CLI (##274)

v0.5.0

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.5.0/manifests/install.yaml

Changes since v0.4.2

Bug Fixes

• Rollout deletionTimestamp are not honored (#109)
• status.availableReplicas should not count old stacks (#143)
• Fix Infinitely loop on controller loop (#146)

Enhancements

• Fast rollback in BlueGreen during scale down period #127
• Attach independent scaleDownDelays to older ReplicaSets #145
• Add scaleDownDelayRevisionLimit to limit the number of old ReplicaSets scaled up #129

Experiment CRD

This release of Argo Rollouts introduces the experiment CRD. The experiment CRD allows users to define multiple PodSpec's to run for a specific duration of time. This will help enable the Kayenta use-case where a user will need to start two versions of their application at the same time. Otherwise, the users cannot have an apples-to-apples comparison of these two versions as one will skew as a result of running for a longer period.

v0.4.2

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.4.2/manifests/install.yaml

Changes since v0.4.1

Bug Fixes

• Honor MaxSurge and MaxUnavailable after last step (#141)
• Fix maxSurge maxUnavailable zero check (#135)
• Add .Spec.Replicas if not set in rollouts (#125)

v0.4.1

Quick Start

kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v0.4.1/manifests/install.yaml

Changes since v0.4.0

Bug Fixes

• Workaround K8s inability to properly validate 'items' subfields #114