All notable changes to this project are documented in this file.
Release date: 2024-12-17
This release comes with support for Splunk Observability (formerly SignalFx) as a metrics provider.
For more information on how to write MetricTemplates
for Splunk, please see the
Splunk metrics tutorial.
Starting with this version, Flagger is compatible with the AWS Gateway API Controller.
Both Flagger and the load tester Go dependencies have been updated to fix various CVEs.
- Add Splunk as a metrics provider #1733
- Preserve HTTPRoute annotations injected by AWS Gateway API #1746
- Automate
zz_generated.deepcopy.go
updates with make codegen #1735 - Update dependencies #1744
Release date: 2024-11-26
This release comes with fixes and improvements. There is a new
.spec.analysis.webhooks[].disableTLS
field which disables TLS verification
for that webhook request.
A bug in the Gateway API provider was fixed which could lead to unecessary restarts.
This release is built with Go 1.23. Lastly, all Go dependencies, Alpine and Kubernetes libraries were updated.
- Add validation for
primaryScalerReplicas
field in the CRD #1702 - feat: add
disableTLS
option for webhooks request #1709 - Update dependencies to Kubernetes v1.31.3 #1723
- Update generated client for Kubernetes 1.31 #1725
- Build with Go 1.23 #1726
- Gateway API: Sort header filters to avoid canary restarts #1713
- fix: fix codegen script and update generated code #1724
- fix(helm): podinfo fails to create the hpa object #1721
Release date: 2024-07-30
This release comes with several fixes and improvements. There is a new Keptn
metrics provider that can be used
for flexible grading logic and analysis.
The loadtester chart now supports ServiceAccount annotations and the Flagger
chart now supports specifying honorLabels
for the PodMonitor.
Support for Kuma has been fixed and verified against Kuma 2.7.5. Also, the
Deployment scaling has been updated to use Patch
instead of Update
to avoid
intermittent conflict errors. Furthermore, a potential panic that could be
caused due to Prometheus returning a range vector has been fixed. Also, the
request-duration
inbuilt query for Nginx has been updated to be more accurate.
Lastly, all Go dependencies, Alpine and Kubernetes libraries were updated.
The update to Kubernetes libraries also brings an unwanted side-effect. Due to
a change in upstream Kubernetes, sidecar support is done through a new field,
which may be utilized by other services in your cluster. This would change the
hash calculated by Flagger between runs and trigger an unwanted Canary
analysis. Unfortunately, this is unavoidable. To get around this, users could
set the .spec.suspend
field to be true before updating to this version and
switch it back when they update their application.
- Bumps golang.org/x/net to v0.23.0 #1628
- feat: implement a Keptn metrics provider #1630
- Update dependencies to Kubernetes 1.30 #1638
- loadtester: add support for annotation on service account #1649
- Bump golang.org/x/net to v0.25.0 and other deps. #1653
- Update Go dependencies and Alpine #1656
- Helm - Add podMonitor.honor labels #1676
- kuma: bump e2e version to 2.7.5 #1683
- Release loadtester 0.33.0 #1690
- Bump google.golang.org/grpc from 1.64.0 to 1.64.1 #1675
- Use
Patch
instead ofUpdate
for Deployment scaling #1634 - block panic when prom returns range vector #1637
- Fix removal of empty keys from flagger chart #1657
- doc: fix KEDA doc regarding namespaces #1666
- Fix Nginx request-duration query #1686
Release date: 2024-03-26
This release updates the Istio APIs to v1beta1
and fixes several issues related
to Gloo routing and custom metrics.
Both Flagger and the load tester Go dependencies have been updated to fix various CVEs. Flagger and the load tester are now built with Go 1.22.
- Migrate Istio VirtualService/DestinationRule APIs to
v1beta1
#1602 - Add
omitempty
to CRD statuses to allow better marshalling #1621 - Update dependencies (Go 1.22) #1622
- Update
google.golang.org/protobuf
to v1.33.0 #1614
- Update reconciler to detect change in Gloo upstream spec #1617
- Fix regression bug where query with no metric template returned an error #1611
Release date: 2024-03-06
This release fixes a bug where .spec..progressDeadlineSeconds
wasn't respected and the Canary
was stuck forever waiting for the Deployment to be ready.
Furthermore, the Go dependencies have been updated.
- Update Go dependencies #1607
- Fix broken link in readme #1599
- scheduler: fail canary according to progress deadline #1603
- Actualize link to flux in-depth guide #1606
Release date: 2024-02-07
This release comes with support for canary releases with traffic shifting using Istio TCP routing. For more information on how to enable TCP routing please see the Istio tutorial.
Both Flagger and the load tester Go dependencies have been updated to fix various CVEs. Flagger is now built with Go 1.21 and the container base image has been updated to Alpine 3.19.
- Istio Canary TCP service support #1564
- Update Go dependencies #1595
- Build with Go 1.21 and Alpine 3.19 #1594
- return an error for missing metric templates #1582
- istio: make retry attempts a mandatory field #1571
- fix(pdb): use the full capabilities comparison for PDBs #1511
Release date: 2023-11-30
This release comes with support for Gateway API v1
. Furthermore, following the
deprecation period, support for the v1alpha2
API has been dropped.
A new field .spec.webhooks[].retries
has been added to allow specifying the
number of retry attempts to make if the webhook server returns an unsuccessful
response.
Another new field .spec.service.trafficPolicy.loadBalancer.warmupDurationSeconds
has been added for the corresponding field in Istio's DestinationRule
API.
Lastly, two bugs related to deleting a Canary object with
.spec.revertOnDeletion: true
have been fixed.
- Support Istio DestinationRule WarmupDurationSecs #1540
- feat: Webhook retries #1541
- gatewayapi: add support for
v1
#1557 - Update Go dependencies #1558
- set original node selector value when finalizing service #1537
- controller: wait for canary deployment to be ready before removing finalizers #1552
Release date: 2023-10-04
This release comes with several new features. The Gateway API integration has been significantly improved with support for
- Canary releases with session affinty
- B/G deployments with traffic mirroring
- Filters in the generated
HTTPRoute
(.spec.rules[].filters
)
Most of the Filters are derived from existing fields in the Canary spec like
.spec.service.headers
. To support arbitary request mirroring through the
RequestMirror
filter, a new field .spec.service.mirror
has been introduced.
A new field checksum
has been added to the Canary webhook payload. This field
is computed by hashing the .status.lastAppliedSpec
and
.status.trackedConfigs
. It can be used to distinguish between Canary runs.
Furthermore, the Gloo integration now uses strings for specifying time durations in order to be better compatible with protobuf duration parsing.
Lastly, Kubernetes packages were updated to be on 1.27.
- Update Kubernetes to v1.27 #1506
- gatewayapi: add support for session affinity #1507
- gatewayapi: add support for route rule filters #1512
- Update Linkerd tutorial to use Kubernetes Gateway API #1516
- Add Checksum field to the Webhook payload to distinguish canary runs #1521
- gatewayapi: add support for b/g mirroring #1525
- Update Go dependencies #1528
- chore: fix incorrect canary name on document #1502
- fix: Support for queryParams in canary match condition #880 #1505
- docs: fix error example in deployment strategies #1518
- Change Gloo Duration type to string #1524
Release date: 2023-08-29
This release fixes bugs related to the Canary lifecycle. The
confirm-traffic-increase
webhook is no longer called if the Canary is in the
WaitingPromotion
phase. Furthermore, a bug which caused downtime when
initializing the Canary deployment has been fixed.
Also, a bug in the request-duration
metric for Traefik which assumed the
result to be in milliseconds instead of seconds has been addressed.
The loadtester now also supports running kubectl
commands.
- Helm: Add option to configure honorLabels for serviceMonitor #1442
- Helm: Use PodDisruptionBudget API policy/v1 if available #1476
- podinfo: Update hpa version from autoscaling/v2beta2 to autoscaling/v2 #1477
- Helm: Allow custom labels for servicemonitor #1483
- feat: loadtester support kubectl type #1485
- Update Istio Gateway reference format #1489
- e2e: Update Istio to v1.18 #1492
- add docs for kubectl in loadtester #1494
- fix: typo on "Parase", should be "Parse". #1443
- Fix Traefik request-duration metric #1446
- Fix initial deployment downtime #1451
- Fix FAQ templating format and change reference of $workload to $target. #1456
- Update doc.go #1466
- Avoid running traffic increase hooks when waiting for promotion or promoting #1470
Release date: 2023-07-14
This release adds support for suspending a Canary using .spec.suspend
.
It also fixes a bug where the target deployment gets stuck at 0 replicas
after the Canary has been deleted.
Furthermore, the Canary API has been modified to allow specifying the
HTTPRoute port using .service.gatewayRefs[].port
.
- Helm: Add option to create service and serviceMonitor #1425
- Update Alpine to 3.18 #1426
- Add
spec.suspend
to allow suspending canary #1431 - Add support for istio LEAST_REQUEST destination rule load balancing #1439
- Add gatewayRef port to Canary CRD #1453
- feat: Copy slowStartConfig for Gloo upstreams #1455
- Update Go dependencies #1459
- Resume target scaler during finalization #1429
- Fix panic when annotation of ingress is empty #1437
- Fixing namespace of HelmRepository in installation docs #1458
Release date: 2023-05-10
This release adds support for Linkerd 2.12 and later. Due to changes in Linkerd
the default namespace for Flagger's installation had to be changed from
linkerd
to flagger-system
and the flagger
Deployment is now injected with
the Linkerd proxy. Furthermore, installing Flagger for Linkerd will result in
the creation of an AuthorizationPolicy
that allows access to the Prometheus
instance in the linkerd-viz
namespace. To upgrade your Flagger installation,
please see the below migration guide.
If you use Kustomize, then follow these steps:
-
kubectl delete -n linkerd deploy/flagger
-
kubectl delete -n linkerd serviceaccount flagger
-
If you're on Linkerd >= 2.12, you'll need to install the SMI extension to enable support for
TrafficSplit
s:curl -sL https://linkerd.github.io/linkerd-smi/install | sh linkerd smi install | kubectl apply -f -
-
kubectl apply -k github.com/fluxcd/flagger//kustomize/linkerd
Note: If you're on Linkerd < 2.12, this will report an error about missing CRDs. It is safe to ignore this error.
If you use Helm and are on Linkerd < 2.12, then you can use helm upgrade
to do
a regular upgrade.
If you use Helm and are on Linkerd >= 2.12, then follow these steps:
helm uninstall flagger -n linkerd
- Install the Linkerd SMI extension:
helm repo add l5d-smi https://linkerd.github.io/linkerd-smi helm install linkerd-smi l5d-smi/linkerd-smi -n linkerd-smi --create-namespace
- Install Flagger in the
flagger-system
namespace and create anAuthorizationPolicy
:helm repo update flagger helm install flagger flagger/flagger \ --namespace flagger-system \ --set meshProvider=linkerd \ --set metricsServer=http://prometheus.linkerd-viz:9090 \ --set linkerdAuthPolicy.create=true
Furthermore, a bug which led the confirm-rollout
webhook to be executed at
every step of the Canary instead of only being executed before the canary
Deployment is scaled up, has been fixed.
- Add support for Linkerd 2.13 #1417
- Fix the loadtester install with flux documentation #1384
- Run
confirm-rollout
checks only before scaling up deployment #1414 - e2e: Remove OSM tests #1423
Release date: 2023-04-12
This release fixes a bug related to the lack of updates to the generated
object's metadata according to the metadata specified in spec.service.apex
.
Furthermore, a bug where labels were wrongfully copied over from the canary
deployment to primary deployment when no value was provided for
--include-label-prefix
has been fixed.
This release also makes Flagger compatible with Flux's helm-controller drift
detection.
- build(deps): bump actions/cache from 3.2.5 to 3.3.1 #1385
- helm: Added the option to supply additional volumes #1393
- build(deps): bump actions/setup-go from 3 to 4 #1394
- update Kuma version and docs #1402
- ci: bump k8s to 1.24 and kind to 1.18 #1406
- Helm: Allow configuring deployment
annotations
#1411 - update dependencies #1412
- Enable updates for labels and annotations #1392
- Update flagger-install-with-flux.md #1398
- avoid copying canary labels to primary on promotion #1405
- Disable Flux helm drift detection for managed resources #1408
Release date: 2023-02-21
This release comes with support for template variables for analysis metrics.
A canary analysis metric can reference a set of custom variables with
.spec.analysis.metrics[].templateVariables
. For more info see the docs.
Furthemore, a bug related to Canary releases with session affinity has been
fixed.
- update dependencies #1374
- build(deps): bump golang.org/x/net from 0.4.0 to 0.7.0 #1373
- build(deps): bump fossa-contrib/fossa-action from 1 to 2 #1372
- Allow custom affinities for flagger deployment in helm chart #1371
- Add namespace to namespaced resources in helm chart #1370
- build(deps): bump actions/cache from 3.2.4 to 3.2.5 #1366
- build(deps): bump actions/cache from 3.2.3 to 3.2.4 #1362
- build(deps): bump docker/build-push-action from 3 to 4 #1361
- modify release workflow to publish rc images #1359
- build: Enable SBOM and SLSA Provenance #1356
- Add support for custom variables in metric templates #1355
- docs(readme.md): add additional tutorial #1346
- use regex to match against headers in istio #1364
Release date: 2023-01-26
This release comes with support for setting a different autoscaling
configuration for the primary workload.
The .spec.autoscalerRef.primaryScalerReplicas
is useful in the
situation where the user does not want to scale the canary workload
to the exact same size as the primary, especially when opting for a
canary deployment pattern where only a small portion of traffic is
routed to the canary workload pods.
- Support for overriding primary scaler replicas #1343
- Allow access to Prometheus in OpenShift via SA token #1338
- Update Kubernetes packages to v1.26.1 #1352
Release date: 2022-12-15
This release comes with support for Apache APISIX. For more details see the tutorial.
- [apisix] Implement router interface and observer interface #1281
- Bump stefanprodan/helm-gh-pages from 1.6.0 to 1.7.0 #1326
- Release loadtester v0.28.0 #1328
- Update release docs #1324
Release date: 2022-11-23
This release comes with support Kubernetes Gateway API v1beta1. For more details see the Gateway API Progressive Delivery tutorial.
Please note that starting with this version, the Gateway API v1alpha2 is considered deprecated and will be removed from Flagger after 6 months.
- Updated Gateway API from v1alpha2 to v1beta1 #1319
- Updated Gateway API docs to v1beta1 #1321
- Update dependencies #1322
- docs: Add
linkerd install --crds
to Linkerd tutorial #1316
Release date: 2022-11-16
This release introduces a new deployment strategy combining Canary releases with session affinity for Istio.
Furthermore, it contains a regression fix regarding metadata in alerts introduced in #1275
- Add support for session affinity during weighted routing with Istio #1280
- Fix cluster name inclusion in alerts metadata #1306
- fix(faq): Update FAQ about zero downtime with correct values #1302
Release date: 2022-10-26
This release comes with a fix to Gloo routing when a custom service name id used.
In addition, the Gloo ingress end-to-end testing was updated to Gloo Helm chart v1.12.31.
- fix(gloo): Use correct route table name in case service name was overwritten #1300
Release date: 2022-10-23
Starting with this version, the Flagger release artifacts are published to GitHub Container Registry, and they are signed with Cosign and GitHub ODIC.
OCI artifacts:
ghcr.io/fluxcd/flagger:<version>
multi-arch container imagesghcr.io/fluxcd/flagger-manifest:<version>
Kubernetes manifestsghcr.io/fluxcd/charts/flagger:<version>
Helm charts
To verify an OCI artifact with Cosign:
export COSIGN_EXPERIMENTAL=1
cosign verify ghcr.io/fluxcd/flagger:1.24.0
cosign verify ghcr.io/fluxcd/flagger-manifests:1.24.0
cosign verify ghcr.io/fluxcd/charts/flagger:1.24.0
To deploy Flagger from its OCI artifacts the GitOps way, please see the Flux installation guide.
- docs: Add guide on how to install Flagger with Flux OCI #1294
- ci: Publish signed Helm charts and manifests to GHCR #1293
- ci: Sign release and containers with Cosign and GitHub OIDC #1292
- ci: Adjust GitHub workflow permissions #1286
- docs: Add link to Flux governance document #1286
Release date: 2022-10-20
This release comes with support for Slack bot token authentication.
- alerts: Add support for Slack bot token authentication #1270
- loadtester: logCmdOutput to logger instead of stdout #1267
- helm: Add app.kubernetes.io/version label to chart #1264
- Update Go to 1.19 #1264
- Update Kubernetes packages to v1.25.3 #1283
- Bump Contour to v1.22 in e2e tests #1282
- gatewayapi: Fix reconciliation of nil hostnames #1276
- alerts: Include cluster name in all alerts #1275
Release date: 2022-08-29
This release fixes a bug related scaling up the canary deployment when a reference to an autoscaler is specified.
Furthermore, it contains updates to packages used by the project, including updates to Helm and grpc-health-probe used in the loadtester.
CVEs fixed (originating from dependencies):
- CVE-2022-37434
- CVE-2022-27191
- CVE-2021-33194
- CVE-2021-44716
- CVE-2022-29526
- CVE-2022-1996
- If HPA is set, it uses HPA minReplicas when scaling up the canary #1253
- Release loadtester v0.23.0 #1246
- Add target and script to keep crds in sync #1254
- docs: add knative support to roadmap #1258
- Update dependencies #1259
- Release loadtester v0.24.0 #1261
Release date: 2022-08-01
This minor release fixes a bug related to the use of HPA v2beta2 and updates
the KEDA ScaledObject API to include MetricType
for ScaleTriggers
.
Furthermore, the project has been updated to use Go 1.18 and Alpine 3.16.
- Update KEDA ScaledObject API to include MetricType for Triggers #1241
- Fix fallback logic for HPAv2 to v2beta2 #1242
Release date: 2022-07-11
This release with support for KEDA ScaledObjects as an alternative to HPAs. Check the tutorial to understand it's usage with Flagger.
The .spec.service.appProtocol
field can now be used to specify the appProtocol
of the services that Flagger generates.
In addition, a bug related to the Contour prometheus query for when service name is overwritten
along with a bug related to a Contour HTTPProxy
annotations have been fixed.
Furthermore, the installation guide for Alibaba ServiceMesh has been updated.
- feat: Add an optional
appProtocol
field tospec.service
#1185 - Update Kubernetes packages to v1.24.1 #1208
- charts: Add namespace parameter to parameters table #1210
- Introduce
ScalerReconciler
and refactor HPA reconciliation #1211 - e2e: Update providers and Kubernetes to v1.23 #1212
- Add support for KEDA ScaledObjects as an auto scaler #1216
- include Contour retryOn in the sample canary #1223
- fix contour prom query for when service name is overwritten #1204
- fix contour httproxy annotations overwrite #1205
- Fix primary HPA label reconciliation #1215
- fix: add finalizers to canaries #1219
- typo: boostrap -> bootstrap #1220
- typo: controller #1221
- update guide for flagger on aliyun ASM #1222
- Reintroducing empty check for metric template references. #1224
Release date: 2022-05-06
This release comes with an option to disable cross-namespace references to Kubernetes
custom resources such as AlertProivders
and MetricProviders
. When running Flagger
on multi-tenant environments it is advised to set the -no-cross-namespace-refs=true
flag.
In addition, this version enables Flagger to target Istio and Kuma multi-cluster setups.
When installing Flagger with Helm, the service mesh control plane kubeconfig secret
can be specified using --set controlplane.kubeconfig.secretName
.
- Add flag to disable cross namespace refs to custom resources #1181
- Rename kubeconfig section in helm values #1188
- Update Flagger overview diagram #1187
- Avoid setting owner refs if the service mesh/ingress is on a different cluster #1183
Release date: 2022-04-15
This release comes with improvements to the AppMesh, Contour and Istio integrations.
- AppMesh: Add annotation to enable Envoy access logs #1156
- Contour: Update the httproxy API and enable RetryOn #1164
- Istio: Add destination port when port discovery and delegation are true #1145
- Metrics: Add canary analysis result as Prometheus metrics #1148
- Fix canary rollback behaviour #1171
- Shorten the metric analysis cycle after confirm promotion gate is open #1139
- Fix unit of time in the Istio Grafana dashboard #1162
- Fix the service toggle condition in the podinfo helm chart #1146
Release date: 2022-03-14
This release comes with support for Kubernetes Gateway API v1alpha2. For more details see the Gateway API Progressive Delivery tutorial.
- Add Gateway API as a provider #1108
- Add arm64 support for loadtester #1128
- Restrict source namespaces in flagger-loadtester #1119
- Remove support for Helm v2 in loadtester #1130
- Fix potential canary finalizer duplication #1125
- Use the primary replicas when scaling up the canary (no hpa) #1110
Release date: 2022-02-14
This release comes with a new API field called canaryReadyThreshold
that allows setting the percentage of pods that need to be available
to consider the canary deployment as ready.
Starting with version, the canary deployment labels, annotations and replicas fields are copied to the primary deployment at promotion time.
- Add field
spec.analysis.canaryReadyThreshold
for configuring canary threshold #1102
- Update metadata during subsequent promote #1092
- Set primary deployment
replicas
when autoscaler isn't used #1106 - Update
matchLabels
forTopologySpreadContstraints
in Deployments #1041
- Send warning and error alerts correctly #1105
- Fix for when Prometheus returns NaN #1095
- docs: Fix typo ExternalDNS #1103
Release date: 2022-01-11
This release comes with support for Kuma Service Mesh. For more details see the Kuma Progressive Delivery tutorial.
To differentiate alerts based on the cluster name, you can configure Flagger with the -cluster-name=my-cluster
command flag, or with Helm --set clusterName=my-cluster
.
- Publish a Software Bill of Materials (SBOM) #1094
- Add cluster name to flagger cmd args for altering #1041
Release date: 2021-12-17
This release contains updates to Kubernetes packages (1.23.0), Alpine (3.15) and load tester components.
- Release loadtester v0.21.0 #1083
- Add loadtester image pull secrets to Helm chart #1076
- Update libraries included in the load tester to newer versions #1063 #1080
- Update Kubernetes packages to v1.23.0 #1078
- Update Alpine to 3.15 #1081
- Update Go to v1.17 #1077
Release date: 2021-11-22
This release comes with a new API field called primaryReadyThreshold
that allows setting the percentage of pods that need to be available
to consider the primary deployment as ready.
- Allow configuring threshold for primary #1048
- Append to list of ownerReferences for primary configmaps and secrets #1052
- Prevent Flux from overriding Flagger managed objects #1049
- Add warning in docs about ExternalDNS + Istio configuration #1044
Release date: 2021-10-28
This release comes with support for NGINX ingress canary metrics. The nginx-ingress minimum supported version is now v1.0.2.
Starting with version, Flagger will use the spec.service.apex.annotations
to annotate the generated apex VirtualService, TrafficSplit or HTTPProxy.
Release date: 2021-09-20
This release comes with support for extending the canary analysis with Dynatrace, InfluxDB and Google Cloud Monitoring (Stackdriver) metrics.
- Add Stackdriver metric provider #991
- Add Influxdb metric provider #1012
- Add Dynatrace metric provider #1013
Release date: 2021-08-25
This release comes with support for Open Service Mesh. For more details see the OSM Progressive Delivery tutorial.
Starting with this version, Flagger container images are signed with sigstore/cosign, for more details see the Flagger cosign docs.
- Support OSM progressive traffic shifting in Flagger #955 #977
- Add support for Google Chat alerts #953
- Sign Flagger container images with cosign #983
- Update Gloo APIs and e2e tests to Gloo v1.8.9 #982
- Update e2e tests to Istio v1.11, Contour v1.18, Linkerd v2.10.2 and NGINX v0.49.0 #979
- Update e2e tests to Traefik to 2.4.9 #960
- Add support for volumes/volumeMounts in loadtester Helm chart #975
- Add extra podLabels options to Flagger Helm Chart #966
- Fix for the http client proxy overriding the default client #943
- Drop deprecated io/ioutil #964
- Remove problematic nulls from Grafana dashboard #952
Release date: 2021-06-17
This release comes with a fix to Flagger when used with Flux v2.
- Update Go to v1.16 and Kubernetes packages to v1.21.1 #940
- Remove the GitOps Toolkit metadata from generated objects #939
Release date: 2021-06-16
This release comes with support for disabling the SSL certificate verification for the Prometheus and Graphite metric providers.
- Add
insecureSkipVerify
option for Prometheus and Graphite #935 - Copy labels from Gloo upstreams #932
- Improve language and correct typos in FAQs docs #925
- Remove Flux GC markers from generated objects #936
- Require SMI TrafficSplit Service and Weight #878
Release date: 2021-06-01
Breaking change: the minimum supported version of Kubernetes is v1.19.0.
This release comes with support for Kubernetes Ingress networking.k8s.io/v1
.
The Ingress from networking.k8s.io/v1beta1
is no longer supported,
affected integrations: NGINX and Skipper ingress controllers.
- Upgrade Ingress to networking.k8s.io/v1 #917
- Update Kubernetes manifests to rbac.authorization.k8s.io/v1 #920
Release date: 2021-05-28
This release comes with support for Graphite metric templates.
- Add Graphite metrics provider #915
- ConfigTracker: Scan envFrom in init-containers #914
- e2e: Update Istio to v1.10 and Contour to v1.15 #914
Release date: 2021-05-14
This release comes with improvements to the Gloo Edge integration.
Starting with this version, Flagger no longer requires Gloo discovery to be enabled.
Flagger generated the Gloo upstream objects on its own and optionally it can use an
existing upstream (specified with .spec.upstreamRef
) as a template.
- Gloo: Create gloo upstreams from non-discovered services #894
- Gloo Upstream Ref for Upstream Config #908
- Adjusted Nginx ingress canary headers on init and promotion #907
Release date: 2021-04-29
This release comes with support for the SMI v1alpha2
and v1alpha3
TrafficSplit APIs.
For SMI compatible service mesh solutions like Open Service Mesh, Consul Connect or Nginx Service Mesh, Prometheus MetricTemplates can be used to implement the request success rate and request duration checks.
The desired SMI version can be set in the Canary object:
apiVersion: flagger.app/v1beta1
kind: Canary
metadata:
name: my-canary
spec:
provider: "smi:v1alpha3" # or "smi:v1alpha2"
- Implement SMI v1alpha2 and v1alpha3 routers #896 #879
- Add alerting HTTP/S proxy option #872
- Add option to mute alerts generated from webhooks #887
- Scale up canary on confirm rollout #878
Release date: 2021-03-23
This release comes with support for manually approving the traffic weight increase.
- Add webhook for manually approving traffic weight increase #849
- Add WaitingPromotion phase to canary status #859
- linkerd: update prometheus URL based on the latest 2.10 changes #845
- docs: update resources to disable mTLS in Istio #843
- docs: updating slack alerting docs to point to legacy slack webhooks #833
- chart: Add pull secret for Prometheus deployment #842
- Update Kubernetes packages to v1.20.4 #857
Release date: 2021-02-26
This release comes with a bug fix to the AppMesh integration when using multiple backends.
- Consolidate logos and add project name logos #829
- chart: add env option to loadtester #821
- chart: Added PodDisruptionBudget for the loadtester #819
- Fix AWS AppMesh issue when providing multiple backends #831
Release date: 2021-02-15
This release comes with support for Kubernetes pod topology spread constraints.
Flagger has a new logo, many thanks to Bianca for designed it.
- Rewrite the primary Pod Topology Spread Constraints based on label selector #806
- Suffix only the podAntiAffinity values that match the deployment name #805
- Check if mandatory secrets/configmaps exist #799
Release date: 2021-01-28
This release comes with support for Kubernetes anti-affinity rules.
- Support for adding
-primary
suffix to Anti-Affinity values #788
- Add missing alerts section to Canary CRD schema #794
Release date: 2021-01-19
This release extends the support for Istio's HTTPMatchRequest
and
comes with a regression bug fix to secrets and configmaps tracking.
- Update HTTPMatchRequest to match Istio's definitions #777
- e2e: Update Istio to v1.8.2 and Contour to v1.11.0 #778
- Add missing TrackedConfig field to Canary status CRD #781
Release date: 2021-01-05
Breaking change: the minimum supported version of Kubernetes is v1.16.0.
This release comes with support for A/B testing using Gloo Edge HTTP headers based routing.
- A/B testing support for Gloo Edge ingress controller #765
- Upgrade the Kubernetes packages to
v1.20.1
and Flagger's CRDs toapiextensions.k8s.io/v1
#772
Release date: 2020-12-22
This is the first release of Flagger under fluxcd organization (CNCF sandbox). Starting with this version, Flagger can be installed on multi-arch Kubernetes clusters (Linux AMD64/ARM64/ARM). The multi-arch image is available on GitHub Container Registry at ghcr.io/fluxcd/flagger.
- Publish multi-arch image to GitHub Container Registry #763
- Migrate CI to GitHub Actions #754
- Add e2e tests for label prefix inclusion #762
- Added PodDisruptionBudget to the Flagger Helm chart #749
Release date: 2020-12-09
Fix Istio virtual service delegation
- Fix for VirtualService delegation when analysis is enabled #745
Prevent primary ConfigMaps and Secrets from being pruned by Flux
- Apply label prefix rules for ConfigMaps and Secrets #743
Add support for Traefik ingress controller
- Add Traefik support for progressive traffic shifting with
TraefikService
#736 - Add support for HPA v2beta2 behaviors #740
Add support for custom weights when configuring traffic shifting
- Support AWS App Mesh backends ARN #715
- Add support for Istio VirtualService delegation #715
- Copy labels from canary to primary workloads based on prefix rules #709
Add support for New Relic metrics
- Add New Relic as a metrics provider #691
- Do not promote when not ready on skip analysis #695
Add support for Skipper ingress controller
- Skipper Ingress Controller support #670
- Support per-config configTracker disable via ConfigMap/Secret annotation #671
- Add priorityClassName and securityContext to Helm charts #652 #668
- Update Kubernetes packages to v1.18.8 #672
- Update Istio, Linkerd and Contour e2e tests #661
- Fix O(log n) bug over network in GetTargetConfigs #663
- Fix(grafana): metrics change since Kubernetes 1.16 #663
Add support for App Mesh Gateway GA
- Update App Mesh docs to v1beta2 API #649
- Add threadiness to Flagger helm chart #643
- Add Istio virtual service to loadtester helm chart #643
This is the GA release for Flagger v1.0.0.
The upgrade procedure from 0.x to 1.0 can be found here.
Two new resources were added to the API: MetricTemplate
and AlertProvider
.
The analysis can reference metric templates
to query Prometheus, Datadog and AWS CloudWatch.
Alerting can be configured on a per
canary basis for Slack, MS Teams, Discord and Rocket.
- Implement progressive promotion #593
- istio: Add source labels to analysis matching rules #594
- istio: Add allow origins field to CORS spec #604
- istio: Change builtin metrics to work with Istio telemetry v2 #623
- appmesh: Implement App Mesh v1beta2 timeout #611
- metrics: Check metrics server availability during canary initialization #592
This is a release candidate for Flagger v1.0.0.
The upgrade procedure from 0.x to 1.0 can be found here.
- Add support for AWS AppMesh v1beta2 API #584
- Add support for Contour v1.4 ingress class #588
- Add user-specified labels/annotations to the generated Services #538
- Support compatible Prometheus service #557
- Update e2e tests and packages to Kubernetes v1.18 #549 #576
- pkg/controller: retry canary initialization on conflict #586
This is a release candidate for Flagger v1.0.0.
The upgrade procedure from 0.x to 1.0 can be found here.
Breaking change: the minimum supported version of Kubernetes is v1.14.0.
- Implement NGINX Ingress header regex matching #546
- pkg/router: update ingress API to networking.k8s.io/v1beta1 #534
- loadtester: add return cmd output option #535
- refactoring: finalizer error handling and unit testing #531 #530
- chart: add finalizers to RBAC rules for OpenShift #537
- chart: allow security context to be disabled on OpenShift #543
- chart: add annotations for service account #521
- docs: Add Prometheus Operator tutorial #524
- pkg/controller: avoid status conflicts on initialization #544
- pkg/canary: fix status retry #541
- loadtester: fix timeout errors #539
- pkg/canary/daemonset: fix readiness check #529
- logs: reduce log verbosity and fix typos #540 #526
This is a release candidate for Flagger v1.0.0.
The upgrade procedure from 0.x to 1.0 can be found here.
- Add opt-in finalizers to revert Flagger's mutations on deletion of a canary #495
- e2e: update end-to-end tests to Contour 1.3.0 and Gloo 1.3.14 #519
- build: update Kubernetes packages to 1.17.4 #516
- Preserve node ports on service reconciliation #514
This is a release candidate for Flagger v1.0.0.
The upgrade procedure from 0.x to 1.0 can be found here.
- Make mirror percentage configurable when using Istio traffic shadowing #492
- Add support for running Concord tests with loadtester webhooks #507
- docs: add Istio telemetry v2 upgrade guide #486, update A/B testing tutorial for Istio 1.5 #502, add how to retry a failed release to FAQ #494
- e2e: update end-to-end tests to Istio 1.5 #447 and NGINX Ingress 0.30 #489 #511
- refactoring: error handling #480, scheduler #484 and unit tests #475
- chart: add the log level configuration to Flagger helm chart #506
- Fix nil pointer for the global notifiers #504
This is a release candidate for Flagger v1.0.0.
The upgrade procedure from 0.x to 1.0 can be found here.
Two new resources were added to the API: MetricTemplate
and AlertProvider
.
The analysis can reference metric templates
to query Prometheus, Datadog and AWS CloudWatch.
Alerting can be configured on a per
canary basis for Slack, MS Teams, Discord and Rocket.
- Implement metric templates for Prometheus #419, Datadog #460 and CloudWatch #464
- Implement metric range validation #424
- Add support for targeting DaemonSets #455
- Implement canary alerts and alert providers (Slack, MS Teams, Discord and Rocket) #429
- Add support for Istio multi-cluster #447 #450
- Extend Istio traffic policy #441, add support for header operations #442 and set ingress destination port when multiple ports are discovered #436
- Add support for rollback gating #449
- Allow disabling ConfigMaps and Secrets tracking #425
Adds support for service name configuration and rollback webhook
Adds event dispatching through webhooks
- Update Contour to v1.1 and add Linkerd header #411
- Update Istio e2e to v1.4.3 #407
- Update Kubernetes packages to 1.17 #406
Adds support for Contour ingress controller
- Add support for Contour ingress controller #397
- Add support for Envoy managed by Crossover via SMI #386
- Extend canary target ref to Kubernetes Service kind #372
- Add Prometheus operator PodMonitor template to Helm chart #399
- Update e2e tests to Kubernetes v1.16 #390
Adds support for taking over a running deployment without disruption
- Add initialization phase to Kubernetes router #384
- Add canary controller interface and Kubernetes deployment kind implementation #378
- Skip primary check on skip analysis #380
Adds wrk to load tester tools and the App Mesh gateway chart to Flagger Helm repository
Adds support for exposing canaries outside the cluster using App Mesh Gateway annotations
- Expose canaries on public domains with App Mesh Gateway #358
- Use the specified replicas when scaling up the canary #363
Fixes promql execution and updates the load testing tools
Adds support for A/B Testing and retry policies when using App Mesh
- Implement App Mesh A/B testing based on HTTP headers match conditions #340
- Implement App Mesh HTTP retry policy #338
- Implement metrics server override #342
- Add the app/name label to services and primary deployment #333
- Allow setting Slack and Teams URLs with env vars #334
- Refactor Gloo integration #344
- Generate unique names for App Mesh virtual routers and routes #336
Adds support for canary and blue/green traffic mirroring
Adds support for App Mesh conformance tests and latency metric checks
- Add support for acceptance testing when using App Mesh #322
- Add Kustomize installer for App Mesh #310
- Update Linkerd to v2.5.0 and Prometheus to v2.12.0 #323
- Fix slack/teams notification fields mapping #318
Adds support for confirm-promotion webhooks and blue/green deployments when using a service mesh
- Canary promotion improvements to avoid dropping in-flight requests #310
- Update end-to-end tests to Kubernetes v1.15.3 and Istio 1.3.0 #306
- Skip primary check for App Mesh #315
Adds support for NGINX custom annotations and Helm v3 acceptance testing
- Add annotations prefix for NGINX ingresses #293
- Add wide columns in CRD #289
- loadtester: implement Helm v3 test command #296
- loadtester: add gRPC health check to load tester image #295
- loadtester: fix tests error logging #286
Adds support for tillerless helm tests and protobuf health checking
- loadtester: add support for tillerless helm #280
- loadtester: add support for protobuf health checking #280
- Set HTTP listeners for AppMesh virtual routers #272
Fixes multi-port support for Istio
- Fix port discovery for multiple port services #267
- Update e2e testing to Istio v1.2.3, Gloo v0.18.8 and NGINX ingress chart v1.12.1 #268
Fixes Blue/Green style deployments for Kubernetes and Linkerd providers
- Fix Blue/Green metrics provider and add e2e tests #261
Adds support for manual gating and pausing/resuming an ongoing analysis
- Implement confirm rollout gate, hook and API #251
- Refactor canary change detection and status #240
- Implement finalising state #257
- Add gRPC load testing tool #248
- Due to the status sub-resource changes in #240,
when upgrading Flagger the canaries status phase will be reset to
Initialized
- Upgrading Flagger with Helm will fail due to Helm poor support of CRDs, see workaround
Adds support for Linkerd (SMI Traffic Split API), MS Teams notifications and HA mode with leader election
Adds support for running Blue/Green deployments without a service mesh or ingress controller
- Allow blue/green deployments without a service mesh provider #211
- Add the service mesh provider to the canary spec #217
- Allow multi-port services and implement port discovery #207
- Add FAQ page to docs website
- Switch to go modules in CI #218
- Update e2e testing to Kubernetes Kind 0.3.0 and Istio 1.2.0
- Update the primary HPA on canary promotion #216
Adds support for customising the Istio traffic policy in the canary service spec
- Generate Istio destination rules and allow traffic policy customisation #200
- Update Kubernetes packages to 1.14 and use go modules instead of dep #202
Adds support for running acceptance/integration tests with Helm test or Bash Bats using pre-rollout hooks
- Implement Helm and Bash pre-rollout hooks #196
- Fix promoting canary when max weight is not a multiple of step #190
- Add ability to set Prometheus url with custom path without trailing '/' #197
Adds support for Service Mesh Interface and Gloo ingress controller
- Add support for SMI (Istio weighted traffic) #180
- Add support for Gloo ingress controller (weighted traffic) #179
Fixes for Jenkins X deployments (prevent the jx GC from removing the primary instance)
- Do not copy labels from canary to primary deployment #178
- Add NGINX ingress controller e2e and unit tests #176
Fixes for custom metrics checks and NGINX Prometheus queries
- Fix promql queries for custom checks and NGINX #174
Adds support for NGINX ingress controller
- Add support for nginx ingress controller (weighted traffic and A/B testing) #170
- Add Prometheus add-on to Flagger Helm chart for App Mesh and NGINX 79b3370
- Fix duplicate hosts Istio error when using wildcards #162
Adds support for SuperGloo
- Supergloo support for canary deployment (weighted traffic) #151
Move Flagger and the load tester container images to Docker Hub
- Add Bash Automated Testing System support to Flagger tester for running acceptance tests as pre-rollout hooks
Adds pre/post rollout webhooks
- Add
pre-rollout
andpost-rollout
webhook types #147
- Set default
mesh
Istio gateway only if no gateway is specified #141
Adds support for App Mesh
- Reconcile Kubernetes ClusterIP services #122
Allows A/B testing scenarios where instead of weighted routing, the traffic is split between the primary and canary based on HTTP headers or cookies.
- A/B testing - canary with session affinity #88
- Update the analysis interval when the custom resource changes #91
Adds support for CORS policy and HTTP request headers manipulation
- Refactor the routing management #72 #80
- Fine-grained RBAC #73
- Add option to limit Flagger to a single namespace #78
Adds support for custom metric checks, HTTP timeouts and HTTP retries
- Allow custom promql queries in the canary analysis spec #60
- Add HTTP timeout and retries to canary service spec #62
Allows for HTTPMatchRequests and HTTPRewrite to be customized in the service spec of the canary custom resource.
- Add HTTP match conditions and URI rewrite to the canary service spec #55
- Update virtual service when the canary service spec changes #54 #51
- Run e2e testing on Kubernetes Kind for canary promotion #53
Allows skipping the analysis phase to ship changes directly to production
- Add option to skip the canary analysis #46
- Reject deployment if the pod label selector doesn't match
app: <DEPLOYMENT_NAME>
#43
Track changes in ConfigMaps and Secrets #37
- Promote configmaps and secrets changes from canary to primary
- Detect changes in configmaps and/or secrets and (re)start canary analysis
- Add configs checksum to Canary CRD status
- Create primary configmaps and secrets at bootstrap
- Scan canary volumes and containers for configmaps and secrets
- Copy deployment labels from canary to primary at bootstrap and promotion
Load testing webhook #35
- Add the load tester chart to Flagger Helm repository
- Implement a load test runner based on rakyll/hey
- Log warning when no values are found for Istio metric due to lack of traffic
- Run wekbooks before the metrics checks to avoid failures when using a load tester
Restart canary analysis if revision changes #31
- Drop support for Kubernetes 1.10
- Detect changes during canary analysis and reset advancement
- Add status and additional printer columns to CRD
- Add canary name and namespace to controller structured logs
- Allow canary name to be different to the target name
- Check if multiple canaries have the same target and log error
- Use deep copy when updating Kubernetes objects
- Skip readiness checks if canary analysis has finished
Configurable canary analysis duration #20
- Helm chart: flag
controlLoopInterval
has been removed
- CRD: canaries.flagger.app v1alpha3
- Schedule canary analysis independently based on
canaryAnalysis.interval
- Add analysis interval to Canary CRD (defaults to one minute)
- Make autoscaler (HPA) reference optional
Webhooks #18
- CRD: canaries.flagger.app v1alpha2
- Implement canary external checks based on webhooks HTTP POST calls
- Add webhooks to Canary CRD
- Move docs to gitbook docs.flagger.app
Improve Slack notifications #14
- Add canary analysis metadata to init and start Slack messages
- Add rollback reason to failed canary Slack messages
Canary progress deadline #10
- Rollback canary based on the deployment progress deadline check
- Add progress deadline to Canary CRD (defaults to 10 minutes)
First stable release
- CRD: canaries.flagger.app v1alpha1
- Notifications: post canary events to Slack
- Instrumentation: expose Prometheus metrics for canary status and traffic weight percentage
- Autoscaling: add HPA reference to CRD and create primary HPA at bootstrap
- Bootstrap: create primary deployment, ClusterIP services and Istio virtual service based on CRD spec
Initial semver release
- Implement canary rollback based on failed checks threshold
- Scale up the deployment when canary revision changes
- Add OpenAPI v3 schema validation to Canary CRD
- Use CRD status for canary state persistence
- Add Helm charts for Flagger and Grafana
- Add canary analysis Grafana dashboard