AWS DevOps Blog

Using Custom JSON on AWS OpsWorks Layers

Custom JSON, which has always been available on AWS OpsWorks stacks and deployments, is now also available as a property on layers in stacks using Chef versions 11.10, 12, and 12.2.

In this post I show how you can use custom JSON to adapt a single Chef cookbook to support different use cases on individual layers. To demonstrate, I use the example of a MongoDB setup with multiple shards.

In OpsWorks, each instance belongs to one or more layers, which in turn make up a stack. You use layers to specify details about which Chef cookbooks are run when the instances are set up and configured, among other things. When your stacks have instances that serve different purposes, you use different cookbooks for each.

Sometimes, however, there are only small differences between the layers and they don’t justify using separate cookbooks. For example, when you have a large MongoDB installation with multiple shards, you would have a layer per shard, as shown in the following figure, but your cookbooks wouldn’t necessarily differ.

custom-json-per-layer-1.png

Let’s assume I’m using the community cookbook for MongoDB. I would configure this cookbook using attributes. The attribute for setting the shard name would be node[:mongodb][:shard_name]. But let’s say that I want to set a certain attribute for any deployment to any instance in a given layer. I would use custom JSON to set the that attribute.

When declared on a stack, custom JSON always applies to all instances, no matter which layer they’re in. Custom JSON declared on a deployment is helpful for one-off deployments with special settings; but, the provided custom JSON doesn’t stick to the instances you deploy to, so a subsequent deployment doesn’t know about custom JSON you might have specified in an earlier deployment.

Custom JSON declared on the layer applies to each instance that belongs to that layer. Like custom JSON declared on the stack, it’s permanently stored and applied to all subsequent deployments. So you just need to edit each layer and set the right shard, as shown in the following figure:

custom-json-per-layer-2.png

During a Chef run, OpsWorks makes custom JSON contents available as attributes. That way the settings are available in the MongoDB cookbook and configure the MongoDB server accordingly. For details about using custom JSON content as an attribute, see our documentation.

Custom JSON declared on the deployment overrides custom JSON declared on the stack. Custom JSON declared on the layer sits in between those two. So you can use it on the layer to override stack settings, and on the deployment to override stack or layer settings.

Using custom JSON gives you a way to tweak a setting for all instances in a given layer without having to affect the entire stack, and without having to provide custom JSON for every deployment.