Setup on Nomad has not been tested.

Using Istio in a non-Kubernetes environment involves a few key tasks:

  1. Setting up the Istio control plane with the Istio API server
  2. Adding the Istio sidecar to every instance of a service
  3. Ensuring requests are routed through the sidecars

Setting up the control plane

Istio control plane consists of four main services: Pilot, Mixer, Citadel, and the API server.

API Server

Istio’s API server (based on Kubernetes’ API server) provides key functions such as configuration management and Role-Based Access Control. The API server requires an etcd cluster as a persistent store. See the instructions for setting up the API server.

Local install

For proof of concept purposes, it is possible to install a simple single container API server using the following Docker-compose file:

version: '2'
          - etcd
      - "4001:4001"
      - "2380:2380"
      - "2379:2379"
    command: [

          - apiserver
      - "8080:8080"
    privileged: true
    command: [
               "kube-apiserver", "--etcd-servers", "http://etcd:2379",
               "--service-cluster-ip-range", "",
               "--insecure-port", "8080",
               "-v", "2",
               "--insecure-bind-address", ""

Other Istio components

Debian packages for Istio Pilot, Mixer, and Citadel are available through the Istio release. Alternatively, these components can be run as Docker containers (,, Note that these components are stateless and can be scaled horizontally. Each of these components depends on the Istio API server, which in turn depends on the etcd cluster for persistence. To achieve high availability, each control plane service could be run as a job in Nomad, where the service stanza can be used to describe the desired properties of the control plane services.

Adding sidecars to service instances

Each instance of a service in an application must be accompanied by the Istio sidecar. Depending on the unit of your installation (Docker containers, VM, bare metal nodes), the Istio sidecar needs to be installed into these components. For example, if your infrastructure uses VMs, the Istio sidecar process must be run on each VM that needs to be part of the service mesh.

One way to package the sidecars into a Nomad-based deployment is to add the Istio sidecar process as a task in a task group. A task group is a collection of one or more related tasks that are guaranteed to be colocated on the same host. However, unlike Kubernetes Pods, tasks in a group do not share the same network namespace. Hence, care must be taken to ensure that only one task group is run per host, when using iptables rules to transparently re-route all network traffic via the Istio sidecar. When support for non-transparent proxying (application explicitly talks to the sidecar) is available in Istio, this restriction will no longer apply.

Routing traffic through Istio sidecars

Part of the sidecar installation should involve setting up appropriate IP Table rules to transparently route application’s network traffic through the Istio sidecars. The IP table script to setup such forwarding can be found in the here.

This script must be executed before starting the application or the sidecar process.

See also

Quick Start instructions to setup the Istio service mesh with Docker Compose.