Single-Host Install with kind and Helm
The root installer in elchi-archive brings the full Elchi stack up on a single Ubuntu 24.04 VM with no existing Kubernetes cluster. It auto-provisions everything — Docker, kind, kubectl and Helm — creates a local kind cluster, and installs the elchi-stack Helm chart into it. It is the fastest way to get a complete evaluation environment on one machine.
Under the hood this runs the same elchi-stack Helm chart as the production Kubernetes install — it just provides the cluster (a local kind cluster) for you. For a real Kubernetes deployment, use the Helm platform install against your own cluster.
When to use it
| Path | Use when |
|---|---|
| kind + Helm (this) | Evaluation or a single VM with no existing Kubernetes cluster |
| Helm | You already run Kubernetes (production) |
| Docker Swarm | Containers on one or more hosts without Kubernetes |
| Bare-metal / systemd | No container runtime — plain systemd services |
Prerequisites
- Ubuntu 24.04 (minimal install). The script warns on other distros/versions.
- Root or sudo privileges and internet connectivity.
- ~15 GB free disk (20 GB recommended) — Docker images, the kind cluster, and the Elchi containers.
Install
Clone the repo and run the root install.sh with your address and the port to expose:
git clone https://github.com/CloudNativeWorks/elchi-archive.git
cd elchi-archive
./install.sh <mainAddress> <port>
For example:
./install.sh elchi.example.com 30080
./install.sh 192.168.1.100 30080
mainAddress— the DNS name or IP clients use to reach Elchi.port— the port to expose the UI on. It becomes the gateway's NodePort, so it must be between 30000 and 32767; the installer refuses anything else. It is mapped from the kind cluster to the host.
What it does
The installer runs through a fixed sequence:
- Validates the OS (expects Ubuntu 24.04) and checks sudo access + disk space.
- Installs the required tools if missing: Docker (from the official Docker apt repo), kubectl, kind
v0.20.0, Helm, and network utilities. - Creates the
elchi-clusterkind cluster — one control-plane + two worker nodes — with your<port>and30001/30002mapped from container to host, and waits for all nodes to be Ready. - Adds the Elchi Helm repo (
https://charts.elchi.io/) and installs theelchi-stackchart into theelchi-stacknamespace, passingglobal.mainAddress,global.port(the gateway's http NodePort), the https NodePort (30001, or30002when your port is 30001) and thestandardstorage class for every volume. - Verifies the pods and services and prints a summary.
When it finishes, the UI is reachable at http://<mainAddress>:<port> (and with a self-signed certificate at https://<mainAddress>:30001). Connect edge hosts with --host=<mainAddress> --port=<port> --tls=false; the installer prints this line too. Pods may still be starting in the background:
kubectl get pods -n elchi-stack -w
Teardown
The root uninstall.sh removes the kind cluster and the tools it installed, while preserving Docker itself and other system packages:
./uninstall.sh
It deletes the elchi-cluster kind cluster, cleans up the Elchi/kind/mongo/victoriametrics/envoy Docker images, and removes the kubectl, kind and helm binaries. To remove only the cluster and keep the tooling, run kind delete cluster --name elchi-cluster instead.
Next steps
- Log in and create your first configuration — see the Quickstart.
- Moving to production Kubernetes — see the Helm platform install.