We would like to watch if some change happens in ConfigMap and/or Secret; then perform a rolling upgrade on relevant DeploymentConfig, Deployment, Daemonset, Statefulset and Rollout

Reloader can watch changes in ConfigMap and Secret and do rolling upgrades on Pods with their associated DeploymentConfigs, Deployments, Daemonsets Statefulsets and Rollouts.

Reloader is available in two different versions:

1. Open Source Version
2. Enterprise Version, which includes:
   • SLA (Service Level Agreement) for support and unique requests
   • Slack support
   • Certified images

Contact sales@stakater.com for info about Reloader Enterprise.

Reloader is compatible with Kubernetes >= 1.19

You have a Deployment called foo and a ConfigMap and/or a Secret either mounted as a volume or defined as a environment variable. The ConfigMap and Secret can be named whatever, but for the sake of this example, lets refer to the ConfigMap as foo-configmap and the secret as foo-secret.

Add the annotation to the main metadata of your Deployment. By default this would be reloader.stakater.com/auto.

kind: Deployment
metadata:
  name: foo
  annotations:
    reloader.stakater.com/auto: "true"
spec:
  template:
    metadata:

This will discover deploymentconfigs/deployments/daemonsets/statefulset/rollouts automatically where foo-configmap or foo-secret is being used either via environment variable or from volume mount. And it will perform rolling upgrade on related pods when foo-configmap or foo-secretare updated.

You can filter it by the type of monitored resource and use typed versions of auto annotation. If you want to discover changes only in mounted Secrets and ignore changes in ConfigMaps, add secret.reloader.stakater.com/auto annotation instead. Analogously, you can use configmap.reloader.stakater.com/auto annotation to look for changes in mounted ConfigMap, changes in any of mounted Secrets will not trigger a rolling upgrade on related pods.

You can also restrict this discovery to only ConfigMap or Secret objects that are tagged with a special annotation. To take advantage of that, annotate your deploymentconfigs/deployments/daemonsets/statefulset/rollouts like this:

kind: Deployment
metadata:
  annotations:
    reloader.stakater.com/search: "true"
spec:
  template:

and Reloader will trigger the rolling upgrade upon modification of any ConfigMap or Secret annotated like this:

kind: ConfigMap
metadata:
  annotations:
    reloader.stakater.com/match: "true"
data:
  key: value

provided the secret/configmap is being used in an environment variable, or a volume mount.

Please note that reloader.stakater.com/search and reloader.stakater.com/auto do not work together. If you have the reloader.stakater.com/auto: "true" annotation on your deployment, then it will always restart upon a change in configmaps or secrets it uses, regardless of whether they have the reloader.stakater.com/match: "true" annotation or not.

Similarly, reloader.stakater.com/auto and its typed version (secret.reloader.stakater.com/auto or configmap.reloader.stakater.com/auto) do not work together. If you have both annotations in your deployment, then only one of them needs to be true to trigger the restart. For example, having both reloader.stakater.com/auto: "true" and secret.reloader.stakater.com/auto: "false" or both reloader.stakater.com/auto: "false" and secret.reloader.stakater.com/auto: "true" will restart upon a change in a secret it uses.

We can also specify a specific configmap or secret which would trigger rolling upgrade only upon change in our specified configmap or secret, this way, it will not trigger rolling upgrade upon changes in all configmaps or secrets used in a deploymentconfig, deployment, daemonset, statefulset or rollout. To do this either set the auto annotation to "false" (reloader.stakater.com/auto: "false") or remove it altogether, and use annotations for Configmap or Secret.

It's also possible to enable auto reloading for all resources, by setting the --auto-reload-all flag. In this case, all resources that do not have the auto annotation (or its typed version) set to "false", will be reloaded automatically when their ConfigMaps or Secrets are updated. Notice that setting the auto annotation to an undefined value counts as false as-well.

To perform rolling upgrade when change happens only on specific configmaps use below annotation.

For a Deployment called foo have a ConfigMap called foo-configmap. Then add this annotation to main metadata of your Deployment

kind: Deployment
metadata:
  annotations:
    configmap.reloader.stakater.com/reload: "foo-configmap"
spec:
  template:
    metadata:

Use comma separated list to define multiple configmaps.

kind: Deployment
metadata:
  annotations:
    configmap.reloader.stakater.com/reload: "foo-configmap,bar-configmap,baz-configmap"
spec:
  template: 
    metadata:

To perform rolling upgrade when change happens only on specific secrets use below annotation.

For a Deployment called foo have a Secret called foo-secret. Then add this annotation to main metadata of your Deployment

kind: Deployment
metadata:
  annotations:
    secret.reloader.stakater.com/reload: "foo-secret"
spec:
  template: 
    metadata:

Use comma separated list to define multiple secrets.

kind: Deployment
metadata:
  annotations:
    secret.reloader.stakater.com/reload: "foo-secret,bar-secret,baz-secret"
spec:
  template: 
    metadata:

• Reloader also supports sealed-secrets. Here are the steps to use sealed-secrets with Reloader.
• For rollouts Reloader simply triggers a change is up to you how you configure the rollout strategy.
• reloader.stakater.com/auto: "true" will only reload the pod, if the configmap or secret is used (as a volume mount or as an env) in DeploymentConfigs/Deployment/Daemonsets/Statefulsets
• secret.reloader.stakater.com/reload or configmap.reloader.stakater.com/reload annotation will reload the pod upon changes in specified configmap or secret, irrespective of the usage of configmap or secret.
• you may override the auto annotation with the --auto-annotation flag
• you may override the secret typed auto annotation with the --secret-auto-annotation flag
• you may override the configmap typed auto annotation with the --configmap-auto-annotation flag
• you may override the search annotation with the --auto-search-annotation flag and the match annotation with the --search-match-annotation flag
• you may override the configmap annotation with the --configmap-annotation flag
• you may override the secret annotation with the --secret-annotation flag
• you may want to prevent watching certain namespaces with the --namespaces-to-ignore flag
• you may want to watch only a set of namespaces with certain labels by using the --namespace-selector flag
• you may want to watch only a set of secrets/configmaps with certain labels by using the --resource-label-selector flag
• you may want to prevent watching certain resources with the --resources-to-ignore flag
• you can configure logging in JSON format with the --log-format=json option
• you can configure the "reload strategy" with the --reload-strategy=<strategy-name> option (details below)

Reloader supports multiple "reload" strategies for performing rolling upgrades to resources. The following list describes them:

• env-vars: When a tracked configMap/secret is updated, this strategy attaches a Reloader specific environment variable to any containers referencing the changed configMap or secret on the owning resource (e.g., Deployment, StatefulSet, etc.). This strategy can be specified with the --reload-strategy=env-vars argument. Note: This is the default reload strategy.
• annotations: When a tracked configMap/secret is updated, this strategy attaches a reloader.stakater.com/last-reloaded-from pod template annotation on the owning resource (e.g., Deployment, StatefulSet, etc.). This strategy is useful when using resource syncing tools like ArgoCD, since it will not cause these tools to detect configuration drift after a resource is reloaded. Note: Since the attached pod template annotation only tracks the last reload source, this strategy will reload any tracked resource should its configMap or secret be deleted and recreated. This strategy can be specified with the --reload-strategy=annotations argument.

You can deploy Reloader by following methods:

You can apply vanilla manifests by changing RELEASE-NAME placeholder provided in manifest with a proper value and apply it by running the command given below:

kubectl apply -f https://raw.githubusercontent.com/stakater/Reloader/master/deployments/kubernetes/reloader.yaml

By default, Reloader gets deployed in default namespace and watches changes secrets and configmaps in all namespaces. Additionally, in the default Reloader deployment, the following resource limits and requests are set:

resources:
  limits:
    cpu: 150m
    memory: 512Mi
  requests:
    cpu: 10m
    memory: 128Mi

Reloader can be configured to ignore the resources secrets and configmaps by passing the following arguments (spec.template.spec.containers.args) to its container:

Note: At one time only one of these resource can be ignored, trying to do it will cause error in Reloader. Workaround for ignoring both resources is by scaling down the Reloader pods to 0.

Reloader can be configured to only watch secrets/configmaps with one or more labels using the --resource-label-selector parameter. Supported operators are !, in, notin, ==, =, !=, if no operator is found the 'exists' operator is inferred (i.e. key only). Additional examples of these selectors can be found in the Kubernetes Docs.

Note: The old : delimited key value mappings are deprecated and if provided will be translated to key=value. Likewise, if a wildcard value is provided (e.g. key:*) it will be translated to the standalone key which checks for key existence.

These selectors can be combined, for example with:

--resource-label-selector=reloader=enabled,key-exists,another-label in (value1,value2,value3)

Only configmaps or secrets labeled like the following will be watched:

kind: ConfigMap
apiVersion: v1
metadata:
  labels:
    reloader: enabled
    key-exists: yes
    another-label: value1

Reloader can be configured to only watch namespaces labeled with one or more labels using the --namespace-selector parameter. Supported operators are !, in, notin, ==, =, !=, if no operator is found the 'exists' operator is inferred (i.e. key only). Additional examples of these selectors can be found in the Kubernetes Docs.

Note: The old : delimited key value mappings are deprecated and if provided will be translated to key=value. Likewise, if a wildcard value is provided (e.g. key:*) it will be translated to the standalone key which checks for key existence.

These selectors can be combined, for example with:

--namespace-selector=reloader=enabled,test=true

Only namespaces labeled as below would be watched and eligible for reloads:

kind: Namespace
apiVersion: v1
metadata:
  labels:
    reloader: enabled
    test: true

You can also apply the vanilla manifests by running the following command

kubectl apply -k https://github.com/stakater/Reloader/deployments/kubernetes

Similarly to vanilla manifests get deployed in default namespace and watches changes secrets and configmaps in all namespaces.

You can write your own kustomization.yaml using ours as a 'base' and write patches to tweak the configuration.

apiVersion: kustomize.config.k8s.io/v1beta1
kind: Kustomization

resources:
  - https://github.com/stakater/Reloader/deployments/kubernetes

namespace: reloader

Alternatively if you have configured helm on your cluster, you can add Reloader to helm from our public chart repository and deploy it via helm using below-mentioned commands. Follow this guide, in case you have trouble migrating Reloader from Helm2 to Helm3.

helm repo add stakater https://stakater.github.io/stakater-charts

helm repo update

helm install stakater/reloader # For helm3 add --generate-name flag or set the release name

helm install {{RELEASE_NAME}} stakater/reloader -n {{NAMESPACE}} --set reloader.watchGlobally=false # By default, Reloader watches in all namespaces. To watch in single namespace, set watchGlobally=false

helm install stakater/reloader --set reloader.watchGlobally=false --namespace test --generate-name # Install Reloader in `test` namespace which will only watch `Deployments`, `Daemonsets` `Statefulsets` and `Rollouts` in `test` namespace.

helm uninstall {{RELEASE_NAME}} -n {{NAMESPACE}}

• Both namespaceSelector & resourceLabelSelector can be used together. If they are then both conditions must be met for the configmap or secret to be eligible to trigger reload events. (e.g. If a configMap matches resourceLabelSelector but namespaceSelector does not match the namespace the configmap is in, it will be ignored).
• At one time only one of the resources ignoreConfigMaps or ignoreSecrets can be ignored, trying to do both will cause error in helm template compilation
• Reloading of OpenShift (DeploymentConfig) and/or Argo Rollouts has to be enabled explicitly because it might not be always possible to use it on a cluster with restricted permissions
• isOpenShift Recent versions of OpenShift (tested on 4.13.3) require the specified user to be in an uid range which is dynamically assigned by the namespace. The solution is to unset the runAsUser variable via deployment.securityContext.runAsUser=null and let OpenShift assign it at install
• reloadOnCreate controls how Reloader handles secrets being added to the cache for the first time. If reloadOnCreate is set to true:
  1. Configmaps/secrets being added to the cache will cause Reloader to perform a rolling update of the associated workload
  2. When applications are deployed for the first time, Reloader will perform a rolling update of the associated workload
  3. If you are running Reloader in HA mode all workloads will have a rolling update performed when a new leader is elected
• reloadOnDelete controls how Reloader handles secrets being deleted. If reloadOnDelete is set to true:
  1. Configmaps/secrets being deleted will cause Reloader to perform a rolling update of the associated workload
• serviceMonitor will be removed in future releases of Reloader in favour of Pod monitor
• If reloadOnCreate is set to false:
  1. Updates to configmaps/secrets that occur while there is no leader will not be picked up by the new leader until a subsequent update of the configmap/secret occurs
  2. In the worst case the window in which there can be no leader is 15s as this is the LeaseDuration
• If reloadOnDelete is set to false:
  1. Deleting of configmaps/secrets has no effect to pods that references these resources.
• By default, reloadOnCreate, reloadOnDelete and syncAfterRestart are all set to false. All need to be enabled explicitly

The Reloader documentation can be viewed from the doc site. The doc source is in the docs folder.

File a GitHub issue.

Join and talk to us on Slack for discussing Reloader:

Please use the issue tracker to report any bugs or file feature requests.

1. Deploy Reloader
2. Run okteto up to activate your development container
3. make build
4. ./Reloader

PRs are welcome. In general, we follow the "fork-and-pull" Git workflow:

1. Fork the repo on GitHub
2. Clone the project to your own machine
3. Commit changes to your own branch
4. Push your work back up to your fork
5. Submit a Pull request so that we can review your changes

NOTE: Be sure to merge the latest from "upstream" before making a pull request!

Repository GitHub releases: As requested by the community in issue 685, Reloader is now based on a manual release process. Releases are no longer done on every merged PR to the main branch, but manually on request.

To make a GitHub release:

1. Code owners create a release branch release-vX.Y.Z
2. Code owners run a dispatch mode workflow to automatically generate version and manifests on the release branch
3. A PR is created to bump the image version on the release branch, example: PR-798
4. Code owners create a GitHub release with tag vX.Y.Z and target branch release-vX.Y.Z, which triggers creation of images

Repository git tagging: Push to the main branch will create a merge-image and merge-tag named merge-${{ github.event.number }}, for example merge-800 when pull request number 800 is merged.

Helm chart versioning: The Reloader Helm chart is maintained in this repository. The Helm chart has its own semantic versioning. Helm charts and code releases are separate artifacts and separately versioned. Manifest making strategy relies on Kustomize. The Reloader Helm chart manages the two artifacts with these two fields:

• appVersion points to released Reloader application image version listed on the releases page
• version sets the Reloader Helm chart version

Helm chart will be released to the chart registry whenever files in deployments/kubernetes/chart/reloader/** change on the main branch.

View the releases page to see what has changed in each release.

Apache2 © Stakater

Reloader is maintained by Stakater. Like it? Please let us know at hello@stakater.com

See our other projects or contact us in case of professional services and queries on hello@stakater.com

• ConfigmapController; We documented here why we re-created Reloader