Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/mikeschinkel/packer-vars

Repo to illustrate a pain when using HashiCorp's Packer for at least one use-case
https://github.com/mikeschinkel/packer-vars

packer packer-builder packer-template packer-templates warnings

Last synced: about 1 month ago
JSON representation

Repo to illustrate a pain when using HashiCorp's Packer for at least one use-case

Awesome Lists containing this project

README 

        
# Packer Vars Example 



## Or, why does it have to be this hard?



This repo was created to provide an example to illustrate my comments on this Github issue for @nywilken who so graciously offered to understand the use-case and possibly even address the difficulty of my use-case.



## Use-case

Basically the use-case did the following, **each with its own `.pkr.hcl` file**. This process assumes an 'ova-builder` VM on VMware vCenter that will have been created by a different Packer template.



| Step      | Description                                                                  |

|-----------|------------------------------------------------------------------------------|

| `setup`   | Ensures the 'ova-builder` VM is running and the vCenter has required config. |

| `clean`   | Ensures the VM is in a clean state, e.g. prior artifacts deleted.            |

| `app`     | Build a GoLang app on the VM.                                                |

| `iso`     | Builds a Linux ISO containing the GoLang App.                                |

| `upload`  | Uploads the build ISO to a vCenter Datastore.                                |

| `install` | Installs the ISO to create a new 'App' VM on vCenter.                        |

| `export`  | Exports the App VM back down to 'ova-builder' as an OVA file.                |

| `jfrog`   | Uploads the resulant OVA to jFrog Artifactory.                               |



The above is even a simplification but is hopefully detailed enough to give a clear picture of the process.



## Example

Aside from the code being proprietary and owned by a former client it would be far too complex to serve as a clarifying example, so I create the simplest example I could to illustrate the problem.



This example has two (2) steps: `step1` and `step2` and their respective Packer templates are `./step1/step1.pkr.hcl` and `./step2/step2.pkr.hcl`.  The example also has a shared `/vars.json` which contains shared vars for logging into SSH and one var for each step, `step1_var=foo` and `step2_var=bar`, respectively. 



All these steps do is they both use the null builder to reach into a Linux machine via SSH and then run an inline script echoing their variable's value.



### Running the working example

To see this example work:



1. Make sure you have Packer installed _(of course)_,

2. Clone this repo to a macOS or Linux machine.

3. Ensure you are in the `working` branch, which is the repo's default.

4. Change the `ssh_*` properties in `./vars.json` to point to a host, user and password of a computer or VM you can SSH into, and then  

4. Run `make`



## The Problem

The problem is that Packer requires you to declare `step2_var` in `./step1/step1.pkr.hcl` even though /step1/step1.pkr.hcl` never references or otherwise uses `step2_var`, — as the following table illustrates:



| Step    | Packer Template         | Template Uses
property in
`./vars.json` | Template does
NOT use: |           

|---------|-------------------------|-----------------------------------------------|---------------------------|

| `step1` | `./step1/step1.pkr.hcl` | `step1_var=foo`                               | `step2_var`               |

| `step2` | `./step2/step2.pkr.hcl` | `step2_var=bar`                               | `step1_var`               |



### Running the "warnings" example



To experience the problem this repo is trying to illustrate:



1. Make sure performed the steps for the `working` example first.

2. Checkout the `warnings` branch.

2. Run `make`



You should get the following output; notice the **warnings**:



```

packer build -force -var-file="./vars.json" "step1"

Warning: Undefined variable



A "step2_var" variable was set but was not found in known variables. To declare

variable "step2_var", place this block in one of your .pkr files, such as

variables.pkr.hcl



null.step2: output will be in this color.



==> null.step2: Using SSH communicator to connect: iso-builder.local

==> null.step2: Waiting for SSH to become available...

==> null.step2: Connected to SSH!

==> null.step2: Provisioning with shell script: /var/folders/fg/1dfmwyrx3wxbhj5lbdqpt7bw0000gn/T/packer-shell1383820945

    null.step2: We are running step 'foo'

Build 'null.step2' finished after 658 milliseconds 319 microseconds.



==> Wait completed after 658 milliseconds 394 microseconds



==> Builds finished. The artifacts of successful builds are:

--> null.step2: Did not export anything. This is the null builder

packer build -force -var-file="./vars.json" "step2"

Warning: Undefined variable



A "step1_var" variable was set but was not found in known variables. To declare

variable "step1_var", place this block in one of your .pkr files, such as

variables.pkr.hcl



null.step2: output will be in this color.



==> null.step2: Using SSH communicator to connect: iso-builder.local

==> null.step2: Waiting for SSH to become available...

==> null.step2: Connected to SSH!

==> null.step2: Provisioning with shell script: /var/folders/fg/1dfmwyrx3wxbhj5lbdqpt7bw0000gn/T/packer-shell2044896123

    null.step2: We are running step 'bar'

Build 'null.step2' finished after 373 milliseconds 507 microseconds.



==> Wait completed after 373 milliseconds 568 microseconds



==> Builds finished. The artifacts of successful builds are:

--> null.step2: Did not export anything. This is the null builder



```



## The Desired Solution



Simply provide a command-line switch that will suppress those warnings.



## Epilogue



I tested this in Packer `1.18.3` but I _swear_ that earlier versions of Packer threw errors and failed to buiild instead of throwing warnings. Or at least that is what I remember. 



So a warning is not as bad as an error, and if it never threw an error and always provided a warning then, as they say, my bad. 



But still, a command-line switch to suppress would be much appreciated.