github.com/kaixiang/packer@v0.5.2-0.20140114230416-1f5786b0d7f1/website/source/intro/getting-started/provision.html.markdown (about) 1 --- 2 layout: "intro" 3 page_title: "Provision" 4 prev_url: "/intro/getting-started/build-image.html" 5 next_url: "/intro/getting-started/parallel-builds.html" 6 next_title: "Parallel Builds" 7 --- 8 9 # Provision 10 11 In the previous page of this guide, you created your first image with 12 Packer. The image you just built, however, was basically just a repackaging 13 of a previously existing base AMI. The real utility of Packer comes from 14 being able to install and configure software into the images as well. 15 This stage is also known as the _provision_ step. Packer fully supports 16 automated provisioning in order to install software onto the machines prior 17 to turning them into images. 18 19 In this section, we're going to complete our image by installing 20 Redis on it. This way, the image we end up building actually contains 21 Redis pre-installed. Although Redis is a small, simple example, this should 22 give you an idea of what it may be like to install many more packages into 23 the image. 24 25 Historically, pre-baked images have been frowned upon because changing 26 them has been so tedious and slow. Because Packer is completely automated, 27 including provisioning, images can be changed quickly and integrated with 28 modern configuration management tools such as Chef or Puppet. 29 30 ## Configuring Provisioners 31 32 Provisioners are configured as part of the template. We'll use the built-in 33 shell provisioner that comes with Packer to install Redis. Modify the 34 `example.json` template we made previously and add the following. We'll 35 explain the various parts of the new configuration following the code 36 block below. 37 38 <pre class="prettyprint"> 39 { 40 "variables": [...], 41 "builders": [...], 42 43 "provisioners": [{ 44 "type": "shell", 45 "inline": [ 46 "sleep 30", 47 "sudo apt-get update", 48 "sudo apt-get install -y redis-server" 49 ] 50 }] 51 } 52 </pre> 53 54 <div class="alert alert-block alert-info"> 55 <strong>Note:</strong> The <code>sleep 30</code> in the example above is 56 very important. Because Packer is able to detect and SSH into the instance 57 as soon as SSH is available, Ubuntu actually doesn't get proper amounts 58 of time to initialize. The sleep makes sure that the OS properly initializes. 59 </div> 60 61 Hopefully it is obvious, but the `builders` section shouldn't actually 62 contain "...", it should be the contents setup in the previous page 63 of the getting started guide. 64 65 To configure the provisioners, we add a new section `provisioners` to the 66 template, alongside the `builders` configuration. The provisioners section 67 is an array of provisioners to run. If multiple provisioners are specified, they 68 are run in the order given. 69 70 By default, each provisioner is run for every builder defined. So if we had 71 two builders defined in our template, such as both Amazon and DigitalOcean, then 72 the shell script would run as part of both builds. There are ways to restrict 73 provisioners to certain builds, but it is outside the scope of this getting 74 started guide. It is covered in more detail in the complete 75 [documentation](/docs). 76 77 The one provisioner we defined has a type of `shell`. This provisioner 78 ships with Packer and runs shell scripts on the running machine. In our 79 case, we specify two inline commands to run in order to install Redis. 80 81 ## Build 82 83 With the provisioner configured, give it a pass once again through 84 `packer validate` to verify everything is okay, then build it using 85 `packer build example.json`. The output should look similar to when you 86 built your first image, except this time there will be a new step where 87 the provisioning is run. 88 89 The output from the provisioner is too verbose to include in this 90 guide, since it contains all the output from the shell scripts. But you 91 should see Redis successfully install. After that, Packer once again 92 turns the machine into an AMI. 93 94 If you were to launch this AMI, Redis would be pre-installed. Cool! 95 96 This is just a basic example. In a real world use case, you may be provisioning 97 an image with the entire stack necessary to run your application. Or maybe 98 just the web stack so that you can have an image for web servers pre-built. 99 This saves tons of time later as you launch these images since everything 100 is pre-installed. Additionally, since everything is pre-installed, you 101 can test the images as they're built and know that when they go into 102 production, they'll be functional.