Using this plugin is a bad idea. It means you've allowed your serverless service to grow in to something huge.
This plugin migrates CloudFormation resources in to nested stacks in order to work around the 500 resource limit.
There are built-in migration strategies that can be turned on or off as well as defining your own custom migrations. It is a good idea to select the best strategy for your needs from the start because the only reliable method of changing strategy later on is to recreate the deployment from scratch. You configure this in your serverless.yml (defaults shown):
custom:
splitStacks:
perFunction: false
perType: true
perGroupFunction: falseThis splits resources off in to a nested stack dedicated to the associated Lambda function. This defaults to off in 1.x but will switch to enabled by default in 2.x
This moves resources in to a nested stack for the given resource type. If Per Lambda is enabled, it takes precedence over Per Type.
This splits resources off in to a nested stack dedicated to a set of Lambda functions and associated resources. If Per Lambda or Per Type is enabled, it takes precedence over Per Lambda Group. In order to control the number of nested stacks, following configurations are needed:
custom:
splitStacks:
nestedStackCount: 20 # Controls the number of created nested stacks
perFunction: false
perType: false
perGroupFunction: trueOnce set, the nestedStackCount configuration should never be changed because the only reliable method of changing it later on is to recreate the deployment from scratch.
In order to avoid API rate limit errors, it is possible to configure the plugin in 2 different ways:
- Set nested stacks to depend on each others.
- Set resources in the nested stack to depend on each others.
This feature comes with a 2 new configurations, stackConcurrency and resourceConcurrency :
custom:
splitStacks:
perFunction: true
perType: false
perGroupFunction: false
stackConcurrency: 5 # Controls if enabled and how much stacks are deployed in parallel. Disabled if absent.
resourceConcurrency: 10 # Controls how much resources are deployed in parallel. Disabled if absent.This plugin is not a substitute for fine-grained services - try to limit the size of your service. This plugin has a hard limit of 200 sub-stacks and does not try to create any kind of tree of nested stacks.
If you create a file in the root of your Serverless project called stacks-map.js this plugin will load it.
This file can customize migrations, either by exporting a simple map of resource type to migration, or a function that can have whatever logic you want.
module.exports = {
'AWS::DynamoDB::Table': { destination: 'Dynamodb' }
}module.exports = (resource, logicalId) => {
if (logicalId.startsWith("Foo")) return { destination: 'Foo' };
// Falls back to default
};You can also point to your custom splitter from the custom block in your serverless file:
custom:
splitStacks:
custom: path/to/your/splitter.js
Be careful when introducing any customizations to default config. Many kind of resources (as e.g. DynamoDB tables) cannot be freely moved between CloudFormation stacks (that can only be achieved via full removal and recreation of the stage)
Custom migrations can specify { force: true } to force the migration of an existing resource in to a new stack. BE CAREFUL. This will cause a resource to be deleted and recreated. It may not even work if CloudFormation tries to create the new one before deleting the old one and they have a name or some other unique property that cannot have two resources existing at the same time. It can also mean a small window of downtime during this period, for example as an AWS::Lambda::Permission is deleted/recreated calls may be denied until IAM sorts things out.
This plugin makes use of the proxy-agent library, which reads environmental varaibles for configuration. To avoid conflicts with existing deployments, it is not used automatically, but instead needs to be enabled via serverless config:
custom:
splitStacks:
proxyAgent: true