Nightscaler is an application designed to annotate Kubernetes namespaces based on Pub/Sub messages. It serves as an extension to kube-downscaler, providing additional functionality for managing forced-uptime periods within namespaces.
Kubernetes clusters often utilize kube-downscaler to scale down namespaces and conserve resources. However, this scaling-down process can render services within a namespace temporarily inaccessible. Nightscaler addresses this limitation by enabling the annotation of forced-uptime periods, instructing kube-downscaler to scale up the affected namespaces during specific time frames.
- Annotations based on Pub/Sub messages: Nightscaler listens for Pub/Sub
messages and expects specific attributes (
action,namespace,cluster, andduration) to be present in each message. - Forced-uptime annotations: When a relevant Pub/Sub message is received,
nightscaler adds an annotation to the targeted namespace, specifying a
forced-uptime period using the
downscaler/force-uptimekey-value pair. - Integration with kube-downscaler: The annotations created by nightscaler trigger kube-downscaler to scale up the namespace during the annotated forced-uptime period, ensuring the availability of services.
- Kubernetes cluster
- Google Cloud Pub/Sub
- Service Account with Pub/Sub permissions and Workload Identity enabled
- Helm
To be added.
First you need to start the Pub/Sub emulator, to do so have a look at the .hack folder.
export PUBSUB_PROJECT_ID=es-standalone-cb21
export PUBSUB_EMULATOR_HOST=localhost:8085
go run .- Create an interface (e.g., a website) that allows users to trigger actions that require forced-uptime periods.
- Configure the interface to send Pub/Sub messages with the required attributes (action, namespace, cluster, and duration) to the nightscaler application.
- Nightscaler will receive the Pub/Sub messages, process them, and add annotations to the specified namespaces accordingly.
- kube-downscaler will respond to the annotations created by nightscaler, scaling up the namespaces during the forced-uptime periods.
Contributions are welcome! If you have any ideas, suggestions, or bug reports, please open an issue or submit a pull request.