Releases: argoproj/argo-rollouts
Releases · argoproj/argo-rollouts
v1.1.1
Quickstart
kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/releases/download/v1.1.1/install.yamlChanges since v1.1.0
Bug Fixes
- fix: Route traffic to Experiment even if Canary RS not scaled (#1638)
- fix: Wait for all canary pods to come up in TrafficRouting canary before switching traffic (#1663)
- fix: inconsistent status command output (#1433)
- fix: add service delete to argo-rollouts role (#1632)
- fix: Fixed NPE while getting the ReplicaSet labels (#1664)
- fix: reset the progress condition when a pod is restarted (#1649)
- fix: Modify Experiment collision naming from dot-notation to dash (#1646)
- fix: release the dashboard-install.yaml (#1601)
- fix: rollout experiment template changing reference rs template labels. Fixes #1596 (#1597)
- fix: viewcontroller gorouting leak in status and get subcommand (#1584)
v1.1.0
Quickstart
kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/releases/download/v1.1.0/install.yamlNotable Features
- Rollout Notifications
- Dynamic scaling of stable ReplicaSet (dynamicStableScale)
- Automated rollbacks without analysis (progressDeadlineAbort)
- Kustomize Open API Schema
- Rollout Dashboard as a Service
- Control scaleDown during abort (abortScaleDownDelaySeconds)
- Analysis: AWS CloudWatch Metric Provider
- AWS TargetGroup IP Verification
- Weighted Experiment Steps with Traffic Routing
- Istio: Multicluster Support
- Istio: TLS Route Support
- Istio: Multiple VirtualServices
- AnalysisRun GC
- Analysis: Graphite Metric Provider
Upgrade Notes
Difference in scale down behavior during aborts
The v1.1 abortScaleDownDelaySeconds feature now allows users full control over the scaling
behavior of the canary/preview ReplicaSet during an abort. Previously in v1.0, it was not possible
to affect this behavior. As part of this feature, v1.1 also fixes some inconsistencies in behavior
with respect to abort scale down.
The most notable change is that upon an abort, the blue-green preview ReplicaSet in v1.1 will now
scale down 30 seconds after the abort, whereas in v1.0 the preview ReplicaSet was left running
indefinitely (without any option to scale it down). If you prefer the v1.0 behavior, you can set
abortScaleDownDelaySeconds: 0, which will leave the preview ReplicaSet running indefinitely
on abort:
spec:
strategy:
blueGreen:
abortScaleDownDelaySeconds: 0Please read the full
documentation to understand
the differences in canary/preview scaling behavior for aborted Rollouts from v1.0 to v1.1.
Changes since v1.0
Controller
- feat: support dynamic scaling of stable ReplicaSet as inverse of canary weight (#1430)
- fix: promote nil pointer error when there are no steps (#1510)
- feat: support management of multiple Istio VirtualService objects (#1381)
- feat: verify AWS TargetGroup after updating active/stable services (#1348)
- feat: ALB TrafficRouting with experiment step
- feat: TrafficRouting SMI with Experiment Step in Canary (#1351)
- feat: ability to abort an update when exceeding progressDeadlineSeconds (#1397)
- feat: add support for Istio VirtualService spec.tls[] (#1380)
- feat: configurable and more aggressive cleanup of old AnalysisRuns and Experiments (#1342)
- feat: ability to auto-create Services for each template in an Experiment (#1158)
- feat: introduce abortScaleDownDelaySeconds to control scale down of preview/canary upon abort (#1160)
- feat: argo rollout compatibility with emissary and edge stack v2.0 (#1330)
- feat: Add support for Istio multicluster (#1274)
- feat: add workload-ref/generation to rollout (#1198)
- feat: support notifications on rollout events using notifications-engine (#1175)
- chore: add liveness and readiness probe to the install manifests (#1324)
- fix: Nginx ingressClassName passed to canary ingress (#1448)
- fix: canary scaledown event could violate maxUnavailable (#1429)
- fix: analysis runs to wait for all metrics to complete (#1407)
- fix: Promote full did not work against BlueGreen with previewReplicaCount (#1384)
- fix: retarget blue-green previewService before scaling up preview ReplicaSet (#1368)
- fix: zero-value abortScaleDownDelay was not honored with setCanaryScale (#1375)
- fix: abort scaledown stable RS for canary with traffic routing (#1331)
Analysis
- feat: add support for Graphite metrics provider (#1406)
- feat: Support CloudWatch as a metric provider (#1338)
- feat: add default() evaluate helper. allow empty datadog result
- fix: Analysis argument validation (#1412)
Plugin
- feat: create windows version for CLI (#1517)
- feat: provide shell completion. Closes #619 (#1478)
- fix: create analysisrun cmd using template generated name (#1471)
- fix: nil pointer in create analysisrun cmd (#1399)
- fix: lint subcommand for workload ref rollout (#1328)
- fix: undo referenced object for workloadRef rollout (#1275)
Dashboard
- feat: allow selection of namespace in rollout dashboard (#1291)
- fix(ui): UI crashes on rollout view due to undefined status (#1287)
Misc
v1.1.0-rc2
fix: Istio does not switch the traffic. Fix the VS new object creatio… …n when the tls routes nil (#1553) Signed-off-by: Andrii Perenesenko <andrii.perenesenko@gmail.com>
v1.0.7
v1.1.0-rc1
Notable Features
- Rollout Notifications
- Dynamic scaling of stable ReplicaSet (dynamicStableScale)
- Automated rollbacks without analysis (progressDeadlineAbort)
- Kustomize Open API Schema
- Rollout Dashboard as a Service
- Control scaleDown during abort (abortScaleDownDelaySeconds)
- Analysis: AWS CloudWatch Metric Provider
- AWS TargetGroup IP Verification
- Weighted Experiment Steps with Traffic Routing
- Istio: Multicluster Support
- Istio: TLS Route Support
- Istio: Multiple VirtualServices
- AnalysisRun GC
- Analysis: Graphite Metric Provider
Upgrade Notes
Difference in scale down behavior during aborts
The v1.1 abortScaleDownDelaySeconds feature now allows users full control over the scaling
behavior of the canary/preview ReplicaSet during an abort. Previously in v1.0, it was not possible
to affect this behavior. As part of this feature, v1.1 also fixes some inconsistencies in behavior
with respect to abort scale down.
The most notable change is that upon an abort, the blue-green preview ReplicaSet in v1.1 will now
scale down 30 seconds after the abort, whereas in v1.0 the preview ReplicaSet was left running
indefinitely (without any option to scale it down). If you prefer the v1.0 behavior, you can set
abortScaleDownDelaySeconds: 0, which will leave the preview ReplicaSet running indefinitely
on abort:
spec:
strategy:
blueGreen:
abortScaleDownDelaySeconds: 0Please read the full
documentation to understand
the differences in canary/preview scaling behavior for aborted Rollouts from v1.0 to v1.1.
Changes since v1.0
Controller
- feat: support dynamic scaling of stable ReplicaSet as inverse of canary weight (#1430)
- fix: promote nil pointer error when there are no steps (#1510)
- feat: support management of multiple Istio VirtualService objects (#1381)
- feat: verify AWS TargetGroup after updating active/stable services (#1348)
- feat: ALB TrafficRouting with experiment step
- feat: TrafficRouting SMI with Experiment Step in Canary (#1351)
- feat: ability to abort an update when exceeding progressDeadlineSeconds (#1397)
- feat: add support for Istio VirtualService spec.tls[] (#1380)
- feat: configurable and more aggressive cleanup of old AnalysisRuns and Experiments (#1342)
- feat: ability to auto-create Services for each template in an Experiment (#1158)
- feat: introduce abortScaleDownDelaySeconds to control scale down of preview/canary upon abort (#1160)
- feat: argo rollout compatibility with emissary and edge stack v2.0 (#1330)
- feat: Add support for Istio multicluster (#1274)
- feat: add workload-ref/generation to rollout (#1198)
- feat: support notifications on rollout events using notifications-engine (#1175)
- chore: add liveness and readiness probe to the install manifests (#1324)
- fix: Nginx ingressClassName passed to canary ingress (#1448)
- fix: canary scaledown event could violate maxUnavailable (#1429)
- fix: analysis runs to wait for all metrics to complete (#1407)
- fix: Promote full did not work against BlueGreen with previewReplicaCount (#1384)
- fix: retarget blue-green previewService before scaling up preview ReplicaSet (#1368)
- fix: zero-value abortScaleDownDelay was not honored with setCanaryScale (#1375)
- fix: abort scaledown stable RS for canary with traffic routing (#1331)
Analysis
- feat: add support for Graphite metrics provider (#1406)
- feat: Support CloudWatch as a metric provider (#1338)
- fix: Analysis argument validation (#1412)
Plugin
- feat: create windows version for CLI (#1517)
- feat: provide shell completion. Closes #619 (#1478)
- fix: create analysisrun cmd using template generated name (#1471)
- fix: nil pointer in create analysisrun cmd (#1399)
- fix: lint subcommand for workload ref rollout (#1328)
- fix: undo referenced object for workloadRef rollout (#1275)
Dashboard
- feat: allow selection of namespace in rollout dashboard (#1291)
- fix(ui): UI crashes on rollout view due to undefined status (#1287)
Misc
v1.0.6
Quickstart
kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/releases/download/v1.0.6/install.yamlChanges since v1.0.4
Bug Fixes
- fix: replica count for new deployment (#1449)
- fix: Nginx ingressClassName passed to canary ingress (#1448)
- fix: Analysis argument validation (#1412)
- fix: retarget blue-green previewService before scaling up preview ReplicaSet (#1368)
- fix: analysis runs to wait for all metrics to complete (#1407)
- fix: canary scaledown event could violate maxUnavailable (#1429)
- chore: release workflow docker build context should use local path and not git context (#1388)
- chore: github release action was using incorect docker cache (#1387)
v1.0.4
Quickstart
kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/releases/download/v1.0.4/install.yamlChanges since v1.0.3
Controller
- fix: Promote full did not work against BlueGreen with previewReplicaCount
v1.0.3
Quickstart
kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/releases/download/v1.0.3/install.yamlChanges since v1.0.2
Controller
- fix: nil pointer dereference when reconciling paused blue-green rollout (#1378)
- fix: Abort rollout doesn't remove all canary pods for setCanaryScale (#1352)
- fix: unsolicited rollout after upgrade from v0.10->v1.0 when pod was using service account (#1367)
- fix: default replica before resolving workloadRef (#1304)
v1.0.2
Quickstart
kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/releases/download/v1.0.2/install.yamlChanges since v1.0.1
Controller
- feat: allow VirtualService HTTPRoute to be inferred if there is single route (#1273)
- fix: rollout paused longer than progressDeadlineSeconds would briefly degrade (#1268)
- fix: controller would drop fields when updating DestinationRules (#1253)
- fix: the wrong panel title on the sample dashboard (#1260)
- fix: analysis with multiple metrics (#1261)
- fix: Mitigate the bug where items are re-added constantly to the workqueue. #1193 (#1243)
- fix: workload rollout spec is invalid template is not empty (#1224)
- fix: Fix error check in validation for AnalysisTemplates not found (#1249)
- fix: make function call consistent with otherRSs definition (#1171)
Plugin
- fix: avoid using root user in plugin container (#1256)
v1.0.1
Quickstart
kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/releases/download/v1.0.1/install.yaml