How negashev/aldaas works
Project type and purpose
• This is an event-driven Kubernetes application (“App with Large Data as a Service”) that automates provisioning of ephemeral, production-like database clones for developers and CI/CD pipelines. It uses Rook/Ceph snapshots, Argo Events, and Argo Workflows under a Helm chart, and is deployed via Rancher Fleet.
File-structure analysis
• .github/workflows: CI to build & publish the tunnel Docker image.
• chart/: Helm chart defining Kubernetes resources (ConfigMaps, CronJobs, Argo Sensors/Sources, RBAC, EventBus, Workflows, Secrets).
• fleet/: Rancher Fleet GitOps manifests for deploying Argo Events, Argo Workflows, Ceph (Rook), and monitoring.
• tunnel/: Light client image (Dockerfile + aldaas.sh) acting as a proxy between user CLI and the Argo control plane.
Key architectural insights from the README
• Core components: Rook-Ceph with VolumeSnapshots, Argo Events (with webhook), Argo Workflows, Helm-deployed control-plane, and a lightweight client proxy.
• The user leverages aldaas.sh + argo CLI to trigger snapshot workflows and bind ephemeral database credentials.
Main components
a. Developer / CI/CD client
b. aldaas tunnel (Docker container)
c. Argo Events:
• EventSource (webhook listener)
• EventBus
• Sensor (triggers on events)
d. Argo Workflow Controller & Server
e. Rook-Ceph storage cluster (VolumeSnapshotClass & snapshots)
f. Kubernetes cluster (namespaces, PVCs, ephemeral database Pods)
g. Rancher Fleet (GitOps: deployments of Argo, Ceph, monitoring)
h. Helm chart (defines all control-plane resources)
Interactions and data flows
1. Client runs aldaas.sh → sends HTTP event to Argo EventSource (via webhook ingress).
2. EventSource publishes to EventBus.
3. Sensor watches EventBus → spawns an Argo Workflow.
4. Workflow steps:
ii. Create a new PVC from that snapshot
iii. Deploy a temporary database Pod bound to the PVC
iv. Generate connection credentials & expose a Service
5. Tunnel proxy forwards client’s DB traffic to the ephemeral Service.
6. (Optionally) CronJob from Helm chart cleans up unused snapshots & PVCs.
Architectural patterns & design principles
• Event-driven, declarative control-plane on Kubernetes.
• Separation of concerns:
• Control plane (Helm chart + Fleet)
• Data plane (Rook/Ceph, PVCs, DB Pods)
• Client plane (tunnel, CLI)
• GitOps deployment (Rancher Fleet) for infra consistency.
• Reuse of Kubernetes primitives (CustomResources, VolumeSnapshots, CronJobs).
Essential elements
• Event flow: Client → EventSource → EventBus → Sensor → Workflow
• Workflow steps hitting the CSI driver & creating DB pods
• Tunnel proxy layer connecting client to ephemeral DB Service
• GitOps loop: Fleet applies fleet/ manifests
• Helm chart role: templating all Kubernetes control-plane resources