docker
/
docs
mirror of https://github.com/docker/docs.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Document UCP architecture and install plan 

Documented install plan, included feedback from team.
Signed-off-by: Joao Fernandes <joao.fernandes@docker.com>
This commit is contained in:
Joao Fernandes 2016-02-12 18:34:46 -08:00
parent 113d8e7a67
commit 85257db9c6
2 changed files with 191 additions and 709 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

BIN
images/ucp-architecture.png Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 56 KiB

872
plan-production-install.md
View File

@ -1,8 +1,8 @@
<!--[metadata]> <!--[metadata]>
+++ +++
title = "Plan a production installation" title = "Plan a production installation"
description = "Plan a production installation" description = "Learn about the Docker Universal Control Plane architecture, and the requirements to install it on production."
keywords = ["tbd, tbd"] keywords = ["docker, ucp, install, checklist"]
[menu.main] [menu.main]
parent="mn_ucp" parent="mn_ucp"
weight=-90 weight=-90
@ -11,766 +11,248 @@ weight=-90
# Plan a production installation # Plan a production installation
Intro 1-2 paras, page purpose, intended user, list steps if page is tutorial. This article gives you an overview of the Docker Universal Control Plane (UCP)
architecture and its components. Here you'll also find all the information
you need to plan a UCP installation on a production environment.
The following topics are covered:
## What is installed * [UCP architecture](#ucp-architecture)
* [The UCP installation](#the-ucp-installation)
* [Understand your installation options](#understand-your-installation-options)
* [Security Considerations](#security-considerations)
* [Docker Engine Configuration](#docker-engine-configuration)
* [Installation checklist](#installation-checklist)
tbd ## UCP architecture
## About the bootstrapper Docker UCP is a cluster with multiple nodes, each running the
Commercially Supported (CS) Docker Engine.
In this cluster, there are 3 different types of nodes:
tbd * [UCP controller node](#UCP-controller-node): the node that handles user
requests,
* [UCP replica nodes](#ucp-replica-nodes): replicas of the controller node
that can take its place if it fails,
* [UCP nodes](#ucp-nodes): the nodes that run your own containers.
### UCP Component architecture ![](images/ucp-architecture.png)
tbd #### UCP controller node
#### controller This is the node that handles user requests to UCP. Users interact with UCP
using a web browser, or a CLI client.
tbd Below is a list of the containers that are deployed to the controller node,
when running the `ucp install` command:
#### nodes | Container name | Description |
|:------------------|:------------------------------------------------------------------------------------------|
| ucp-controller | Manages UCP resources, admin configuration, and user commands. |
| ucp-swarm-manager | Accepts requests the from UCP controller. |
| ucp-kv | Internal node discovery, cluster configuration, and support for HA. |
| ucp-ca | Allow external systems to reach UCP. Proxy for TLS. |
| ucp-swarm-ca | Allow external systems to reach UCP, Swarm, and Engine, using credentials. Proxy for TLS. |
tbd
#### images #### UCP replica nodes
tbd Docker UCP has support for high availability. You can configure replica nodes
to stand by, and be ready to take the place of the controller if it fails.
#### containers A cluster with N controller and replica nodes can only tolerate (N-1)/2 node
failures. So be sure to set up at least two replicas, when installing UCP for
production.
tbd Below is a list of the containers that are deployed to the controller node,
when running the `ucp join --replica` command:
### CS Docker Engine | Container name | Description |
|:------------------|:--------------------------------------------------------------------|
| ucp-controller | Manages UCP resources, admin configuration, and user commands. |
| ucp-swarm-manager | Accepts requests from the UCP controller. |
| ucp-kv | Internal node discovery, cluster configuration, and support for HA. |
| ucp-proxy | Manages TLS and requests from swarm manager. |
| ucp-swarm-join | Heartbeat for Swarm nodes to ensure they are running. |
tbd In UCP v1.0 the controller serves as root CA, and no other nodes are able to
sign certificates. If the controller fails, it might not be possible to add new
nodes to the cluster. [Learn more about high availability](understand_ha.md)
### Docker Swarm #### UCP nodes
tbd These are the nodes that run your own containers. When you deploy
your containers using UCP, it spreads the load across these nodes. You can
configure the strategies used by UCP to deploy the containers across multiple
nodes.
### Key/value Store Below is a list of the containers that are deployed to the controller node,
when running the `ucp join` command:
tbd | Container name | Description |
|:---------------|:------------------------------------------------------|
| ucp-proxy | Manages TLS and requests from swarm manager. |
| ucp-swarm-join | Heartbeat for Swarm nodes to ensure they are running. |
## Understand your installation options
tbd ## The UCP installation
### Evaluation install is a good trial to do
tbd
### AMI or manual installation
tbd
### Default versus the custom installation options
tbd
### Mix panel analytics
tbd
## Network considerations
tbd
### Port numbers
tbd
### Swarm Manager
tbd
### Data Volumes
tbd
### High availability
tbd
### Discovery
tbd
## Security Considerations
tbd
### Certificate authority
tbd
### Controlling user access
tbd
## Docker Engine Configuration
tbd
### Starting the daemon
tbd
### Docker CLI and the certificate bundle
tbd
## Installation checklist
tbd
### Hardware and software requirements
tbd
### IP addresses and fully-qualified domain names
tbd
### Subject alternative names (SANs)
tbd
## Where to go next
<!--[metadata]>
# Docker UCP Quickstart Guide
These instructions explain how to install Docker Universal Control Plane (UCP). A UCP installation consists of an UCP controller and one or more nodes. The same machine can serve as both the controller and the node. These instructions show you how to install both a host and a node. It contains the following sections:
- [Plan your installation](#plan-your-installation)
- [Step 1: Verify you have the prerequisites](#step-1-verify-you-have-the-prerequisites)
- [Step 2: Configure your network for UCP](#step-2-configure-your-network-for-ucp)
- [Step 3: Install Docker CS Engine 1.9](#step-3-install-docker-cs-engine)
- [Step 4: (optional) Create user-named volumes](#step-4-optional-create-user-named-volumes)
- [Step 5: Install the UCP controller](#step-5-install-the-ucp-controller)
- [Step 6: (optional) Add a controller replica to the UCP cluster](#step-6-optional-add-a-controller-replica-to-the-ucp-cluster)
- [Step 7: Add an Engine node to the UCP cluster](#step-7-add-an-engine-node-to-the-ucp-cluster)
- [Step 8: Set up certs for the Docker CLI](#step-8-set-up-certs-for-the-docker-cli)
- [Uninstall](#uninstall)
- [Block Mixpanel analytics](#block-mixpanel-analytics)
- [Installing with your own certificates](#installing-with-your-own-certificates)
- [Where to go next](#where-to-go-next)
**Upgrade Note**: If you have installed UCP 0.6.0 or higher, you can upgrade to UCP 0.7.0 using the [upgrade](reference/upgrade.md) command.
## Plan your installation
The UCP installation consists of using the Docker Engine CLI to run the `ucp` The UCP installation consists of using the Docker Engine CLI to run the `ucp`
tool. The `ucp` tool is an image with subcommands to `install` a controller or tool. The `ucp` tool is a Docker image that can run subcommands like `install`
`join` a node to a UCP controller. The general format of these commands are: to install UCP on a node, or `join` to join a node to a UCP cluster.
| Docker client | `run` command with options | `ucp` image | Subcommand with options | The general format of these commands is:
|---------------|----------------------------|--------------|-------------------------|
| Docker client | run command with options | ucp image | Subcommand with options |
|:--------------|:-------------------------|:-------------|:------------------------|
| `docker` | `run --rm -it` | `docker/ucp` | `install --help` | | `docker` | `run --rm -it` | `docker/ucp` | `install --help` |
| `docker` | `run --rm -it` | `docker/ucp` | `join --help` | | `docker` | `run --rm -it` | `docker/ucp` | `join --help` |
| `docker` | `run --rm -it` | `docker/ucp` | `uninstall --help` | | `docker` | `run --rm -it` | `docker/ucp` | `uninstall --help` |
You can these two subcommands interactively by passing them the `-i` You can use the `ucp` tool interactively, by passing them the `-i` option, or
option or by passing command-line options. This installation guide's steps non-interactively, by passing command-line options.
assume both are run interactively.
To list all the possible subcommands, use: To list all the possible subcommands for the `ucp` tool, use:
```
$ docker run --rm -it docker/ucp --help $ docker run --rm -it docker/ucp --help
```
### Default versus the custom installation options You can also find more information about the `ucp` command at the
[UCP tool reference page](reference/index.md).
This installation guide walks you through The `ucp` tool pulls several images from Docker Hub, and runs them in
containers. The only change the `ucp` tool makes to your system, is
creating certificates under the
[data volumes you create when installing](#data-volumes-used).
* configures data volumes on the controller ## Understand your installation options
* generates a certificate chain on the controller
* creates a Swarm cluster The best way to test-drive UCP is to install and run it on a sandbox.
* loads and launches the appropriate images The default options of the `ucp install` command, make it easy to test UCP.
[Learn how to install UCP on your local machine](evaluation-install.md).
Regardless of how you use the `docker/ucp install` command, the installer When installing UCP for production, you'll want to ensure your installation
supplies some quick default options for both data volumes and the certificate scales and is secure. You'll need to:
authority (CA).
The first time you install, you should build a sandbox environment using * Set up UCP for high availability,
`ucp` defaults. After installing and using this sandbox environment, * Customize some of the ports used,
you can uninstall it and try a custom installation. In a custom installation you * Customize the data volumes used,
can: * Use your own certificates.
* use the high availability feature [Learn how to install UCP in production](production-install.md).
* customize the port used by the UCP web application
* customize the port used by the Swarm manager
* create your own data volumes
* use your own certs
This install documentation describes the default installation and the ## Installation checklist
customization steps. Customize steps are identified with the keyword
(optional). Make sure you skip these steps when doing the default installation
in your sandbox.
### IP addresses and fully-qualified domain names Before installing UCP, make sure you have the following requirements:
The `docker/ucp install` subcommand works to do as much network discovery as it * [Hardware and sofware requirements](#hardware-and-software-requirements)
can. If your network configuration defines fully-qualified domain names (FQDN) * [Ports used](#ports-used)
on each host, the installer uses this information to obtain accurate host * [IP addresses and fully-qualified domain names](#ip-addresses-and-fully-qualified-domain-names)
addresses. If you haven't set up your network with FQDNs and the installer * [Subject alternative names (SANs)](#subject-alternative-names-sans)
discovery fails, you must supply the installer an accessible host address either * [Data volumes used](#data-volumes-used)
interactively or using the `--host-address` option.
If you are using a cloud provider such as AWS or Digital Ocean, you may need to ### Hardware and software requirements
allocate a private network for your UCP installation. You can use this network
as long as the controller and nodes can communicate via
their private IPs. If the private IPs do not support communication among the
UCP cluster, using public IPs or full-qualified domain names are required. For
more information about what ports and protocols are required see [Step 2: Configure your network for UCP](#step-2-configure-your-network-for-ucp).
### Subject alternative names (SANs) You can install UCP on-premises or on a cloud provider such AWS or Digital
Ocean. To install UCP, all nodes must have:
Further, UCP requires that all clients, including the Docker Engine, use a Swarm
TLS certificate chain signed by the UCP Swarm Root CA. You provide the
certificate system with subject alternative names or SANs. The SANs are used to
set up individual "leaf certificates."
For the controller and each node, you must specify at least one SAN; you can
specify more. You can pass the `--san` values to the boostrapper's `install` or
`join`. A SAN value can be the pubic IP address and/or fully-qualified domain
name.
If you are using a cloud provider and specified private IPs for the host address
values, consider whether you need to access your cluster through a public
network as well as the private network space. If the answer is yes, your SAN
values should contain both the public IPs or full-qualified domain names (FQDN) and the private network IPs.
### Mixpanel analytics
The UCP BETA program makes use of Mixpanel to collect analytics. This feature
collects data on your usage of UCP and returns it to Docker. The information is
entirely anonymous and does not identify your Company or users. Currently, you
cannot turn the collection off but you can block the outgoing messaging. Later
in this documentation [Block Mixpanel analytics](#block-mixpanel-analytics)
explains how.
### About these installation instructions
These installation instructions were written using the Ubuntu 14.0.3 operating
system. The file paths and commands used in the instructions are specific to
Ubuntu 14.0.3. If you are installing on another operating system, the steps are
the same but your commands and paths may differ.
The Ubuntu system used to write these instructions was an AWS EC2 instance. The
instance was running in a in an AWS virtual private network. Elastic IPs were
configured for all the hosts.
## Step 1: Verify you have the prerequisites
You can install UCP on your network or on a cloud provider such AWS or Digital
Ocean. To install, the controller and the nodes require a minimum of 1.50 GB of memory. You can run any of these supported operating systems:
* 1.50 GB of RAM
* 3.00 GB of available disk space
* One of the supported operating systems installed:
* RHEL 7.0, 7.1 * RHEL 7.0, 7.1
* Ubuntu 14.04 LTS * Ubuntu 14.04 LTS
* CentOS 7.1 * CentOS 7.1
* Kernel version 3.10 or higher
* CS Docker Engine installed
Your system must have a 3.16.0 kernel or higher. If you don't have the proper kernel installed, the `docker/ucp install` command returns this error. If the `ucp` tool detects that your system doesn't have these
requirements, it displays a warning message.
``` ### Ports used
INFO[0000] Verifying your system is compatible with UCP
FATA[0000] Your kernel version 3.13.0 is too old. UCP requires at least version 3.16.0 for all features to work. To proceed with an old kernel use the '--old-kernel' flag
```
If you proceed with the '--old-kernel' flag, you cannot use the Docker custom networking features. The following ports are using on UCP installation:
Installing Docker UCP requires that you first install the CS Docker Engine v1.9 | Hosts | Direction | Port | Purpose |
on both the controller and the nodes. The CS Engine can be installed manually or |:----------------------------|:---------:|:--------------------|:------------------------------------------------------------|
from an image if your cloud provider support its. These instructions assume you | controller, replicas | in | 443 (configurable) | web app and CLI client access to UCP. |
are installing both UCP and Engine manually. | controller, replicas | in | 2376 (configurable) | swarm manager accepts requests from UCP controller. |
| controller, replicas, nodes | in | 2375 | heartbeat for nodes, to ensure they are running. |
| controller, replicas, nodes | in | 12376 | proxy for TLS, provides access to UCP, Swarm, and Engine. |
| controller, replicas | in | 12379 | internal node configuration, cluster configuration, and HA. |
| controller, replicas | in | 12380 | internal node configuration, cluster configuration, and HA. |
| controller | in | 12381 | proxy for TLS, provides access to UCP. |
| controller | in | 12382 | manages TLS and requests from swarm manager. |
| controller, replicas | out | 443 | send anonymous usage reports to Docker. |
## Step 2: Configure your network for UCP UCP collects anonymous data on the usage of UCP and reports to Docker.
This data is entirely anonymous and does not identify your company or users.
UCP includes Docker Swarm as part of its installation. So, you don't need to You can disable this at any time on the UCP settings screen.
install Docker Swarm. You do need to ensure that the UCP controller and nodes
can communicate across your network. Configure your network making sure to open
the following ports:
<table>
<tr>
<th>Port</th>
<th>Description</th>
</tr>
<tr>
<td><code>443</code> </td>
<td>UCP controller port. Open this port to allow inbound access to the UCP interface or CLI.</td>
</tr>
<tr>
<td><code>2376</code></td>
<td>Allows inbound access from the nodes to the Swarm manager.</td>
</tr>
<tr>
<td><code>12376</code></td>
<td>Open this on secondary nodes to allow the Engine proxy to access the Orca server node.</td>
</tr>
<tr>
<td><code>12379</code>, <code>12380</code></td>
<td>Key Value store </td>
</tr>
<tr>
<td><code>12381</code></td>
<td>Swarm CA service</td>
</tr>
<tr>
<td><code>12382` </td>
<td>UCP CA service</td>
</tr>
</table>
All the communication among the controller, nodes, and key value store is ### IP addresses and fully-qualified domain names
protected by mutual TLS. The UCP installation of Swarm provides and configures
TLS for you automatically.
Finally, you can specify a different port for the Swarm manager if you need to. The `ucp install` command tries to use your network configurations to find
Using a different port is a customization. These instructions assume you are using the default `2376` port. the fully-qualified domain names (FQDN) of your hosts.
## Step 3: Install Docker CS Engine If you haven't set up your network with FQDNs, the installer might not be able
to find the FQDN for the host. If that happens, you'll be prompted for an
accessible host address. You can also use the `--host-address` option to
specify the FQDN to the install command.
The BETA program requires that you install the Docker CS Engine 1.9.0 or above. If you're installing UCP on a cloud provider such as AWS or Digital Ocean,
Follow the instructions for your particular operating system and ensure you might need to create a private network for you UCP installation. In that
you are pointing at the proper repo. case, make sure all nodes of the cluster can communicate using their private
IPs.
Install the Docker CS Engine on both the controller node and each member node. If the nodes in the cluster cannot communicate using the private IPs,
you'll need to use public IPs or Fully Qualified Domain Names.
### RHEL 7.0, 7.1 and CentOS 7.1 [Check the ports used](#ports-used) by UCP.
Use the detailed [Red Hat Linux installation
instructions](https://docs.docker.com/docker-trusted-registry/install/install-csengine/#centos-7-1-rhel-7-0-7-1-yum-based-systems)
### Ubuntu 14.04 LTS
Use the [detailed Ubuntu installation
instructions](https://docs.docker.com/docker-trusted-registry/install/install-csengine/#install-on-ubuntu-14-04-lts)
## Step 4: (optional) Create user-named volumes If you are using a cloud provider such as AWS or Digital Ocean, you may need to
create a private network for your UCP installation. You can use this network
as long as the controller and nodes can communicate via
their private IPs. If the private IPs do not support communication among the
UCP cluster, using public IPs or full-qualified domain names are required. For
more information about what ports and protocols are required see
[Step 2: Configure your network for UCP](#step-2-configure-your-network-for-ucp).
UCP uses named volumes for persistence of user data. By default, the `ucp
install` command creates for you. It uses the default volume driver and flags.
The first time you install, you should skip this step and try it later. Later,
try an install where your take the option to use custom volume driver and create
your own volumes.
If you choose this option, create your volumes prior to installing UCP. The ### Subject alternative names (SANs)
volumes UCP requires are:
All the communication between the nodes is protected by mutual TLS. The UCP
installation configures TLS for you automatically.
For that, UCP requires that all clients, including the Docker Engine, use a
Swarm TLS certificate chain signed by the UCP Swarm Root CA. You provide the
certificate system with subject alternative names or SANs. The SANs are used to
set up individual "leaf certificates."
When using the `ucp install` or `ucp join` commands interactively, you are
prompted for the subject alternative names of the host. This can be a public
IP address, or fully-qualified domain name.
You can also send the `--san` option to specify the SANs for a host.
### Data volumes used
UCP uses named volumes for persistence of data. When installing UCP for
production, you need to create the following volumes:
| Volume name | Data | | Volume name | Data |
|-------------------------|--------------------------------------------------------------------------------------| |:---------------------------|:---------------------------------------------------------------------------------------------------------------|
| `ucp-root-ca` | The certificate and key for the UCP root CA. Do not create this volume if you are using your own certificates. | | ucp-root-ca | The certificate and key for the UCP root CA. Do not create this volume if you are using your own certificates. |
| `ucp-swarm-root-ca` | The certificate and key for the Swarm root CA. | | ucp-swarm-root-ca | The certificate and key for the Swarm root CA. |
| `ucp-server-certs` | The controller certificates for the UCP controllers web server. | | ucp-server-certs | The controller certificates for the UCP controllers web server. |
| `ucp-swarm-node-certs` | The Swarm certificates for the current node (repeated on every node in the cluster). | | ucp-swarm-node-certs | The Swarm certificates for the current node (repeated on every node in the cluster). |
| `ucp-swarm-kv-certs` | The Swarm KV client certificates for the current node (repeated on every node in the cluster). | | ucp-swarm-kv-certs | The Swarm KV client certificates for the current node (repeated on every node in the cluster). |
| `ucp-swarm-controller-certs` | The UCP Controller Swarm client certificates for the current node. | | ucp-swarm-controller-certs | The UCP Controller Swarm client certificates for the current node. |
| `ucp-kv` | Key value store persistence. | | ucp-kv | Key value store persistence. |
## Step 5: Install the UCP controller If you don't create these volumes, the `ucp install` command creates them for
you with the default volume driver and flags.
In this step you install the UCP controller. The controller includes a running
Swarm manager and node as well. Use the following command to pull the
`docker/ucp` image and review the `install` options:
```bash
docker run --rm -it docker/ucp install --help
```
When you install, the script prompts you for the following information:
* a password to use for the UCP `admin` account
* at lease one SAN value which is the accessible IP address or fully-qualified domain name for the controller node
When you have the information you'll be prompted for, do the following to
install:
1. Log into the system where you mean to install the UCP controller.
If you are installing on a cloud provider such as AWS, make sure the instance has a public IP or hostname.
![Open certs](images/ip_cloud_provider.png)
2. Run the `ucp` command interactively.
$ docker run --rm -it -v /var/run/docker.sock:/var/run/docker.sock --name ucp docker/ucp install -i
The command pulls several images and prompts you for the installation values it needs. When it completes, the command prompts you to login into the UCP GUI.
INFO[0053] Login to UCP at https://10.0.0.32:443
3. Enter the address into your browser to view the UCP login screen.
Your browser may warn you about the connection. The warning appears because
the UCP certificate was issued by a built-in certificate authority. Your
actions with the install actually created the certificate. If you are
concerned, the certificate's fingerprint is displayed during install and you
can compare it.
4. Use the Advanced link to proceed to UCP.
The login screen displays.
![](images/login.png)
5. Enter `admin` for the username along with the password you provided to the `install`.
If you didn't enter an admin password, the default password is `orca` After
you enter the correct credentials, the UCP dashboard displays.
![](images/dashboard.png)
The dashboard shows a single node, your controller node.
## Step 6: (optional) Add a controller replica to the UCP cluster
In this optional step, you configure support for UCP's high-availability
feature. You do this by adding one or more UCP *replicas* using the `docker/ucp
join` subcommand. The first time you install, you should skip this optional
step and try it later. Later, try an install where you configure
high-availability.
When adding nodes to your cluster, you decide which nodes you to use as
*replicas* and which nodes are simply for extra capacity. A
replica is a node in your cluster that can act as an additional UCP controller.
Should the primary controller fail, a replica can take over the controller role
for the cluster. If you are trying out the optional HA deployment:
* Configure the controller and replicas before adding additional Engine nodes.
* Configure a minimum of two replicas in addition to the controller.
Repeat the install for each node you want to add. Use the following command to pull the `docker/ucp` image and review the `join` options:
```bash
docker run --rm -it docker/ucp join --help
```
The `docker/ucp join` prompts you for the following information:
* the URL of the UCP controller, for example `https://52.70.188.239`
* the username/password of an UCP administrator account
* at least one SAN value which is an accessible IP address or fully-qualified domain name for node
When you have the information you'll be prompted for, do the following to install:
1. Log into the host where you mean to install the node.
2. Run the `dokcer/ucp join` command.
$ docker run --rm -it -v /var/run/docker.sock:/var/run/docker.sock --name ucp docker/ucp join --replica -i
The `join` pulls several images and prompts you for the installation values
it needs. When it completes, the command notifies you that it is starting
swarm.
INFO[0005] Verifying your system is compatible with UCP
INFO[0011] Sending add host request to UCP server
INFO[0011] Starting local swarm containers
3. Repeat steps 1 thru 3 on the other replicas.
You should configure a minimum of 3 controllers configured, a primary and
two replicas. Never run a cluster with only the primary controller and a
single replica.
4. Login into UCP with your browser and navigate to the **NODES** page.
Simply clicking on the nodes from the Dashboard takes you to the page. The page should display your new nodes.
![](images/nodes.png)
## Step 7: Add an Engine node to the UCP cluster
In this step, you install one or more UCP nodes using the `ucp join` subcommand. Repeat the install for each node you want to add. Use the following command to pull the `docker/ucp` image and review the `join` options:
```bash
docker run --rm -it docker/ucp join --help
```
The `docker/ucp join` command prompts you for the following information:
* the URL of the UCP controller, for example `https://52.70.188.239`
* the username/password of an UCP administrator account
* at least one SAN value which is the actual external, publically-accessible IP address or fully-qualified domain name for node
When you have the information you'll be prompted for, do the following to install:
1. Log into the system where you mean to install the node.
2. Run the `docker/ucp join` command.
$ docker run --rm -it -v /var/run/docker.sock:/var/run/docker.sock --name ucp docker/ucp join -i
The `join` pulls several images and prompts you for the installation values it needs. When it completes, the command notifies you that it is starting Swarm.
INFO[0005] Verifying your system is compatible with UCP
INFO[0011] Sending add host request to UCP server
INFO[0011] Starting local swarm containers
3. Repeat steps 1 thru 2 on the other nodes.
4. Login into UCP with your browser and navigate to the **NODES** page.
Simply clicking on the nodes from the Dashboard takes you to the page. The page should display your new nodes.
![](images/nodes.png)
## Step 8: Set up certs for the Docker CLI
Once you install UCP on a machine, it is a good idea to download a client bundle. The bundle contains the certificates a user needs to run the Docker Engine command line client (`docker`) against the UCP controller and nodes.
You can download the bundle from the UCP interface or using `curl` command. The
bundle is in a `.zip` package. You need `zip` or similar to unzip the file. If you plan to use `curl` you also need JQuery. This is used to pass the `curl`
command an authorization token. Of course, you need to have `curl` installed as well.
### Download the bundle from the UCP interface
1. If you haven't already done so, log into UCP.
2. Choose **ADMIN > Profile** from the right-hand menu.
Any user can download their certificates. So, if you were logged in under a user name such as `davey` the path to download bundle is **davey > Profile**. Since you are logged ins as `ADMIN`, the path is `ADMIN`.
3. Click **Create Client Bundle**.
The browser downloads the `ucp-bundle-admin.zip` file.
### Download the bundle with curl
1. Log into a machine with network access to the UCP controller.
You might log into the controller itself. You could also log into any arbitrary machine able to `ping` the controller.
2. Install the prerequisite `curl`, `zip`, `jq` (JQuery) packages if you haven't already.
On Ubuntu, the installation looks like this:
$ sudo apt-get install zip curl jq
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following extra packages will be installed:
libcurl3
The following NEW packages will be installed:
jq zip
The following packages will be upgraded:
curl libcurl3
----output snipped----
To curl the bundle, you must export your user security token from the UCP controller. You do this in the next step.
3. Create an environment variable to hold your user security token.
AUTHTOKEN=$(curl -sk -d '{"username":"admin","password":"<password>"}' https://<ducp-0 IP>/auth/login | jq -r .auth_token)
4. Curl the client bundle down to your node.
$ curl -k -H "Authorization: Bearer $AUTHTOKEN" https://<ducp-0 IP>/api/clientbundle -o bundle.zip
The browser downloads a `bundle.zip` file.
### Install the certificate bundle
Once you download the bundle, you can unzip and use it.
1. Make sure you have `zip` installed.
$ which unzip
/usr/bin/unzip
If you don't, install it before continuing.
2. Open the folder containing the bundle file.
4. Unzip the file to reveal its contents.
ucp-bundle
├── ca.pem
├── cert.pem
├── cert.pub
├── env.sh
└── key.pem
5. Set up your environment by sourcing the `env.sh` file.
$ source env.sh
6. Use the `docker info` command to get the location of the Swarm managers and engines.
$ docker info
Containers: 9
Images: 9
Role: primary
Strategy: spread
Filters: health, port, dependency, affinity, constraint
Nodes: 1
node1: 192.168.122.7:12376
└ Containers: 9
└ Reserved CPUs: 0 / 1
└ Reserved Memory: 0 B / 2.054 GiB
└ Labels: executiondriver=native-0.2, kernelversion=4.0.9-boot2docker, operatingsystem=Boot2Docker 1.8.1 (TCL 6.3); master : eb5571f - Thu Sep 3 22:18:54 UTC 2015, provider=kvm, storagedriver=aufs
Cluster Managers: 1
192.168.122.7: Healthy
└ Orca Controller: https://192.168.122.7
└ Swarm Manager: tcp://192.168.122.7:3376
└ KV: etcd://192.168.122.7:12379
CPUs: 1
Total Memory: 2.054 GiB
Name: node1
ID: PNLT:MFCO:DDWL:MSLF:YVHU:35Z3:66KM:DFZM:OPBK:D4BQ:EKNT:6DXA
Labels:
com.docker.ucp.license_key=unlicensed
com.docker.ucp.license_max_engines=0
com.docker.ucp.license_expires=EXPIRED
### Client Bundles on Externally Managed CA Configuration
If UCP is configured with an external CA, it will be unable to sign client bundles for non-admin users automatically. It is still possible to manually issue certificates signed by the CA that UCP users can use to interact with UCP via the CLI.
Generate an 2048-bit RSA private key.
```
openssl genrsa -out key.pem 2048
```
Generate a Certificate Signing Request (CSR). The output `cert.csr` should be provided to your organization's CA owner to be signed, with a minimum of client authentication usage.
```
openssl req -new -sha256 -key key.pem -out cert.csr
```
Your CA owner will sign the CSR, and provide `cert.pem` and `ca.pem` files.
Extract the public key from the signed certificate:
```
openssl x509 -pubkey -noout -in cert.pem > cert.pub
```
The contents of cert.pub will then need to be added to your profile. You can add this in the UI by clicking the User Menu in the top right corner, and select profile.
Once you are on the User Profile screen, click the "Add an Existing Public Key" button and provide the contents of cert.pub, along with a memorable label for this bundle.
Now that you have linked the public key to you account, the next step is to configure your CLI. To configure your CLI to use the certificate bundle that you have generated, you will need to export the following environment variables:
```
export DOCKER_TLS_VERIFY=1
export DOCKER_CERT_PATH=$(pwd)
export DOCKER_HOST=tcp://<ucp-hostname>:443
```
## Uninstall
The `docker/ucp uninstall` command removes UCP from the controller and the nodes. The uninstall process does not remove any other containers that are running, except those recognized to be part of UCP. To see the `uninstall` options before you uninstall, use the following:
```bash
docker run --rm -it docker/ucp uninstall --help
```
To uninstall, do the following:
1. Log into the node you want to remove UCP from.
2. Enter the following command to uninstall:
$ docker run --rm -it -v /var/run/docker.sock:/var/run/docker.sock --name ucp docker/ucp uninstall -i
3. Repeat the uninstall on each node making sure to save the controller till last.
## Block Mixpanel analytics
To block the outflow of Mixpanel analytic data to Docker, do the following:
1. Log into the system running the UCP controller.
2. Add a rule to drop the forward to port 80.
$ iptables -I FORWARD -p tcp --dport 80 -j DROP
Reboots unset this iptables chain, so it is a good idea to add this command to the controller's startup script.
## Installing with your own certificates
UCP uses two separate root CAs for access control - one for Swarm, and one for
the UCP controller itself. The dual root certificates supply differentiation
between the Docker remote API access to UCP vs. Swarm. Unlike Docker Engine or
Docker Swarm, UCP implements ACL and audit logging on a per-user basis. Swarm
and the Engine proxies trust only the Swarm Root CA, while the UCP controller
trusts both Root CAs. Admins can access UCP, Swarm and the engines while
normal users are only granted access to UCP.
UCP v1.0 supports user provided externally signed certificates
for the UCP controller. This cert is used by UCP's main management web UI
and the Docker remote API. The remote API is visible to the Docker CLI. In this release, the Swarm Root CA is always managed by UCP.
The external UCP Root CA model supports customers managing their own CA, or
purchasing certs from a commercial CA.
The first time you install, we recommend you skip user-supplied certs and use
the default certificates instead. The default TLS certificate files are placed
on the host filesystem of each Docker Engine in
`/var/lib/docker/discovery_certs/`. Later, do a second install and try the
option to use your own certs.
### Configure user-supplied Certificates
To install UCP with your own external root CA, you create a named volume called
**ucp-server-certs** on the same system where you plan to install the UCP
controller.
1. Log into the machine where you intend to install UCP.
2. If you haven't already done so, create a named volume called **ucp-server-certs**.
3. Ensure the volume's top-level directory contains these files:
<table>
<tr>
<th>File</th>
<th>Description</th>
</tr>
<tr>
<td><code>ca.pem</code></td>
<td>Your Root CA Certificate chain (including any intermediaries).</td>
</tr>
<tr>
<td><code>cert.pem</code></td>
<td>Your signed UCP controller cert.</td>
</tr>
<tr>
<td><code>key.pem</code></td>
<td>Your UCP controller private key.</td>
</tr>
</table>
4. Follow "Step 5" above to install UCP but pass in an additional `--external-ucp-ca` option to the `docker/run install`, for example:
docker run --rm -it \
-v /var/run/docker.sock:/var/run/docker.sock \
...snip...
install -i --external-ucp-ca
## Where to go next ## Where to go next
To learn more you can also investigate: * [Test-drive Docker UCP by installing it on a sandbox](evaluation-install.md)
* [Install Docker UCP on production](production-install.md)
* Read more [about the product](https://www.docker.com/universal-control-plane).
* Visit the UCP forum to [ask questions and get answers](https://forums.docker.com/c/commercial-products/ucpbeta).
* Try the [UCP hands-on lab](https://github.com/docker/ucp_lab).
* [How to use the Docker Client](http://docs.docker.com/reference/commandline/cli/)
* [An overview of Docker Swarm](http://docs.docker.com/swarm/)
<![end-metadata]-->