Installing in an air-gapped environment

Some components of the Elastic Stack require additional configuration and local dependencies in order to deploy in environments without internet access. This guide gives an overview of this setup scenario and helps bridge together existing documentation for individual parts of the stack.

Refer to the section for each Elastic component for air-gapped installation configuration and dependencies in a self-managed Linux environment.

Air-gapped install of Elasticsearch may require additional steps in order to access some of the features. General install and configuration guides are available in the Elasticsearch install documentation.

Specifically:

Air-gapped install of Kibana may require a number of additional services in the local network in order to access some of the features. General install and configuration guides are available in the Kibana install documentation.

Specifically:

Elastic Beats are light-weight data shippers. They do not require any unique setup in the air-gapped scenario. To learn more, refer to the Beats documentation.

Logstash is a versatile data shipping and processing application. It does not require any unique setup in the air-gapped scenario. To learn more, refer to the Logstash documentation.

Air-gapped install of Elastic Agent depends on the Elastic Package Registry and the Elastic Artifact Registry for most use-cases. The agent itself is fairly lightweight and installs dependencies only as required by its configuration. In terms of connections to these dependencies, Elastic Agents need to be able to connect to the Elastic Artifact Registry directly, but Elastic Package Registry connections are handled through Kibana.

Additionally, if the Elastic Agent Elastic Defend integration is used, then access to the Elastic Endpoint Artifact Repository is necessary in order to deploy updates for some of the detection and prevention capabilities.

To learn more about install and configuration, refer to the Elastic Agent install documentation. Make sure to check the requirements specific to running Elastic Agents in an air-gapped environment.

To get a better understanding of how to work with Elastic Agent configuration settings and policies, refer to Appendix D - Agent Integration Guide.

Fleet Server is a required middleware component for any scalable deployment of the Elastic Agent. The air-gapped dependencies of Fleet Server are the same as those of the Elastic Agent.

To learn more about installing Fleet Server, refer to the Fleet Server set up documentation.

Air-gapped setup of the APM server is possible in two ways:

Refer to Connect to Elastic Maps Service in the Kibana documentation to learn how to configure your firewall to connect to Elastic Maps Service, host it locally, or disable it completely.

Detailed install and configuration instructions are available in the Enterprise Search install documentation.

Air-gapped install of the EPR is possible using any OCI-compatible runtime like Podman (a typical choice for RHEL-like Linux systems) or Docker. Links to the official container image and usage guide is available on the Air-gapped environments page in the Fleet and Elastic Agent Guide.

Refer to Appendix A - Elastic Package Registry for additional setup examples.

Air-gapped install of the Elastic Artifact Registry is necessary in order to enable Elastic Agent deployments to perform self-upgrades and install certain components which are needed for some of the data integrations (that is, in addition to what is also retrieved from the EPR). To learn more, refer to Host your own artifact registry for binary downloads in the Fleet and Elastic Agent Guide.

Refer to Appendix B - Elastic Artifact Registry for additional setup examples.

Air-gapped setup of this component is, essentially, identical to the setup of the Elastic Artifact Registry except that different artifacts are served. To learn more, refer to Configure offline endpoints and air-gapped environments in the Elastic Security guide.

Some machine learning features, like natural language processing (NLP), require you to deploy trained models. To learn about deploying machine learning models in an air-gapped environment, refer to:

Detailed install and configuration instructions are available in the Kibana AI Assistants settings documentation.

Setting up air-gapped Kubernetes or OpenShift installs of the Elastic Stack has some unique concerns, but the general dependencies are the same as in the self-managed install case on a regular Linux machine.

The Elastic Kubernetes operator is an additional component in the Kubernetes OpenShift install that, essentially, does a lot of the work in installing, configuring, and updating deployments of the Elastic Stack. For details, refer to the Elastic Cloud on Kubernetes install instructions.

The main requirements are:

The container image can be downloaded from the official Elastic Docker repository, as described in the Fleet and Elastic Agent air-gapped environments documentation.

This container would, ideally, run as a Kubernetes deployment. Refer to Appendix C - EPR Kubernetes Deployment for examples.

A custom container would need to be created following similar instructions to setting up a web server in the self-managed install case. For example, a container file using an NGINX base image could be used to run a build similar to the example described in Appendix B - Elastic Artifact Registry.

Just like the Elastic Artifact Registry. A custom container needs to be created following similar instructions to setting up a web server for the self-managed install case.

Besides the public Elastic container repository, most Elastic Stack container images are also available in Platform One’s Iron Bank.

To install Elastic Cloud Enterprise in an air-gapped environment you’ll need to host your own 1.10. Elastic Package Registry. Refer to the ECE offline install instructions for details.

The following script generates a SystemD service file on a RHEL 8 system in order to run EPR with Podman in a production environment.

#!/usr/bin/env bash

EPR_BIND_ADDRESS="0.0.0.0"
EPR_BIND_PORT="8443"
EPR_TLS_CERT="/etc/elastic/epr/epr.pem"
EPR_TLS_KEY="/etc/elastic/epr/epr-key.pem"
EPR_IMAGE="docker.elastic.co/package-registry/distribution:9.0.0-beta1"

podman create \
  --name "elastic-epr" \
  -p "$EPR_BIND_ADDRESS:$EPR_BIND_PORT:$EPR_BIND_PORT" \
  -v "$EPR_TLS_CERT:/etc/ssl/epr.crt:ro" \
  -v "$EPR_TLS_KEY:/etc/ssl/epr.key:ro" \
  -e "EPR_ADDRESS=0.0.0.0:$EPR_BIND_PORT" \
  -e "EPR_TLS_CERT=/etc/ssl/epr.crt" \
  -e "EPR_TLS_KEY=/etc/ssl/epr.key" \
  "$EPR_IMAGE"

## creates service file in the root directory
# podman generate systemd --new --files --name elastic-epr --restart-policy always

The following is an example of an actual SystemD service file for an EPR, launched as a Podman service.

# container-elastic-epr.service
# autogenerated by Podman 4.1.1
# Wed Oct 19 13:12:33 UTC 2022

[Unit]
Description=Podman container-elastic-epr.service
Documentation=man:podman-generate-systemd(1)
Wants=network-online.target
After=network-online.target
RequiresMountsFor=%t/containers

[Service]
Environment=PODMAN_SYSTEMD_UNIT=%n
Restart=always
TimeoutStopSec=70
ExecStartPre=/bin/rm -f %t/%n.ctr-id
ExecStart=/usr/bin/podman run \
	--cidfile=%t/%n.ctr-id \
	--cgroups=no-conmon \
	--rm \
	--sdnotify=conmon \
	-d \
	--replace \
	--name elastic-epr \
	-p 0.0.0.0:8443:8443 \
	-v /etc/elastic/epr/epr.pem:/etc/ssl/epr.crt:ro \
	-v /etc/elastic/epr/epr-key.pem:/etc/ssl/epr.key:ro \
	-e EPR_ADDRESS=0.0.0.0:8443 \
	-e EPR_TLS_CERT=/etc/ssl/epr.crt \
	-e EPR_TLS_KEY=/etc/ssl/epr.key docker.elastic.co/package-registry/distribution:9.0.0-beta1
ExecStop=/usr/bin/podman stop --ignore --cidfile=%t/%n.ctr-id
ExecStopPost=/usr/bin/podman rm -f --ignore --cidfile=%t/%n.ctr-id
Type=notify
NotifyAccess=all

[Install]
WantedBy=default.target

The following example script downloads artifacts from the internet to be later served as a private Elastic Package Registry.

#!/usr/bin/env bash
set -o nounset -o errexit -o pipefail

STACK_VERSION=9.0.0-beta1
ARTIFACT_DOWNLOADS_BASE_URL=https://artifacts.elastic.co/downloads

DOWNLOAD_BASE_DIR=${DOWNLOAD_BASE_DIR:?"Make sure to set DOWNLOAD_BASE_DIR when running this script"}

COMMON_PACKAGE_PREFIXES="apm-server/apm-server beats/auditbeat/auditbeat beats/elastic-agent/elastic-agent beats/filebeat/filebeat beats/heartbeat/heartbeat beats/metricbeat/metricbeat beats/osquerybeat/osquerybeat beats/packetbeat/packetbeat cloudbeat/cloudbeat endpoint-dev/endpoint-security fleet-server/fleet-server"

WIN_ONLY_PACKAGE_PREFIXES="beats/winlogbeat/winlogbeat"

RPM_PACKAGES="beats/elastic-agent/elastic-agent"
DEB_PACKAGES="beats/elastic-agent/elastic-agent"

function download_packages() {
  local url_suffix="$1"
  local package_prefixes="$2"

  local _url_suffixes="$url_suffix ${url_suffix}.sha512 ${url_suffix}.asc"
  local _pkg_dir=""
  local _dl_url=""

  for _download_prefix in $package_prefixes; do
    for _pkg_url_suffix in $_url_suffixes; do
          _pkg_dir=$(dirname ${DOWNLOAD_BASE_DIR}/${_download_prefix})
          _dl_url="${ARTIFACT_DOWNLOADS_BASE_URL}/${_download_prefix}-${_pkg_url_suffix}"
          (mkdir -p $_pkg_dir && cd $_pkg_dir && curl -O "$_dl_url")
    done
  done
}

# and we download
for _os in linux windows; do
  case "$_os" in
    linux)
      PKG_URL_SUFFIX="${STACK_VERSION}-${_os}-x86_64.tar.gz"
      ;;
    windows)
      PKG_URL_SUFFIX="${STACK_VERSION}-${_os}-x86_64.zip"
      ;;
    *)
      echo "[ERROR] Something happened"
      exit 1
      ;;
  esac

  download_packages "$PKG_URL_SUFFIX" "$COMMON_PACKAGE_PREFIXES"

  if [[ "$_os" = "windows" ]]; then
    download_packages "$PKG_URL_SUFFIX" "$WIN_ONLY_PACKAGE_PREFIXES"
  fi

  if [[ "$_os" = "linux" ]]; then
    download_packages "${STACK_VERSION}-x86_64.rpm" "$RPM_PACKAGES"
    download_packages "${STACK_VERSION}-amd64.deb" "$DEB_PACKAGES"
  fi
done


## selinux tweaks
# semanage fcontext -a -t "httpd_sys_content_t" '/opt/elastic-packages(/.*)?'
# restorecon -Rv /opt/elastic-packages

The following is an example NGINX configuration for running a web server for the Elastic Artifact Registry.

user  nginx;
worker_processes  2;

error_log  /var/log/nginx/error.log notice;
pid        /var/run/nginx.pid;

events {
    worker_connections  1024;
}

http {
    include       /etc/nginx/mime.types;
    default_type  application/octet-stream;

    log_format  main  '$remote_addr - $remote_user [$time_local] "$request" '
                      '$status $body_bytes_sent "$http_referer" '
                      '"$http_user_agent" "$http_x_forwarded_for"';

    access_log          /var/log/nginx/access.log  main;
    sendfile            on;
    keepalive_timeout   65;

    server {
        listen                  9080 default_server;
        server_name             _;
        root                    /opt/elastic-packages;

        location / {

        }
    }

}

The following is a sample EPR Kubernetes deployment YAML file.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: elastic-package-registry
  namespace: default
  labels:
    app: elastic-package-registry
spec:
  replicas: 1
  selector:
    matchLabels:
      app: elastic-package-registry
  template:
    metadata:
      name: elastic-package-registry
      labels:
        app: elastic-package-registry
    spec:
      containers:
        - name: epr
          image: docker.elastic.co/package-registry/distribution:9.0.0-beta1
          ports:
            - containerPort: 8080
              name: http
          livenessProbe:
            tcpSocket:
              port: 8080
            initialDelaySeconds: 20
            periodSeconds: 30
          resources:
            requests:
              cpu: 125m
              memory: 128Mi
            limits:
              cpu: 1000m
              memory: 512Mi
          env:
            - name: EPR_ADDRESS
              value: "0.0.0.0:8080"
---
apiVersion: v1
kind: Service
metadata:
  labels:
    app: elastic-package-registry
  name: elastic-package-registry
spec:
  ports:
  - port: 80
    name: http
    protocol: TCP
    targetPort: http
  selector:
    app: elastic-package-registry

When configuring any integration in Elastic Agent, you need to set up integration settings within whatever policy is ultimately assigned to that agent.

Note the following terms and definitions:

There are three ways to configure Elastic Agent integrations:

Best option for: Manual configuration and users who prefer using a UI over scripting.

Example: Get started with logs and metrics

Agent policies and integration settings can be managed using the Kibana UI. For example, the following shows the configuration of logging for the System integration in an Elastic Agent policy:

Good option for: Declarative configuration and users who need reproducible and automated deployments.

Example: Fleet settings in Kibana

You can have Kibana create Elastic Agent policies on your behalf by adding appropriate configuration parameters in the kibana.yml settings file, these include:

Best option for: Declarative configuration and users who need reproducible and automated deployments in even the trickiest of environments.

Example: See the following.

It is possible to use custom scripts that call the Kibana Fleet API to create or update policies without restarting Kibana, and also allowing for custom error handling and update logic.

At this time, you can refer to the the Kibana Fleet HTTP API documentation, however additional resources from public code repositories should be consulted to capture the full set of configuration options available for a given integration. Specifically, many integrations have configuration options such as inputs and data_streams that are unique.

In particular, the *.yml.hbs templates should be consulted to determine which vars are available for configuring a particular integration using the Kibana Fleet API.