bluesliverx / vals

Helm-like configuration values loader with support for various sources

vals

vals is a tool for managing configuration values and secrets.

It supports various backends including:

• Vault

• AWS SSM Parameter Store

• AWS Secrets Manager

• AWS S3

• GCP Secrets Manager

• SOPS-encrypted files

• Terraform State

• CredHub(Coming soon)

• Use vals eval -f refs.yaml to replace all the refs in the file to actual values and secrets.

• Use vals exec -f env.yaml -- <COMMAND> to populate envvars and execute the command.

• Use vals env -f env.yaml to render envvars that are consumable by eval or a tool like direnv

Usage

• CLI
• Helm
• Go

CLI

vals is a Helm-like configuration "Values" loader with support for various sources and merge strategies

Usage:
  vals [command]

Available Commands:
  eval		Evaluate a JSON/YAML document and replace any template expressions in it and prints the result
  exec		Populates the environment variables and executes the command
  env		Renders environment variables to be consumed by eval or a tool like direnv
  ksdecode	Decode YAML document(s) by converting Secret resources' "data" to "stringData" for use with "vals eval"

Use "vals [command] --help" for more infomation about a command

vals has a collection of providers that each an be referred with a URI scheme looks ref+<TYPE>.

For this example, use the Vault provider.

Let's start by writing some secret value to Vault:

$ vault kv put secret/foo mykey=myvalue

Now input the template of your YAML and refer to vals' Vault provider by using ref+vault in the URI scheme:

$ VAULT_TOKEN=yourtoken VAULT_ADDR=http://127.0.0.1:8200/ \
  echo "foo: ref+vault://secret/data/foo?proto=http#/mykey" | vals eval -f -

Voila! vals, replacing every reference to your secret value in Vault, produces the output looks like:

foo: myvalue

Which is equivalent to that of the following shell script:

VAULT_TOKEN=yourtoken  VAULT_ADDR=http://127.0.0.1:8200/ cat <<EOF
foo: $(vault kv get -format json secret/foo | jq -r .data.data.mykey)
EOF

Save the YAML content to x.vals.yaml and running vals eval -f x.vals.yaml does produce output equivalent to the previous one:

foo: myvalue

Helm

Use value references as Helm Chart values, so that you can feed the helm template output to vals -f - for transforming the refs to secrets.

$ helm template mysql-1.3.2.tgz --set mysqlPassword='ref+vault://secret/data/foo#/mykey' | vals ksdecode -o yaml -f - | tee manifests.yaml
apiVersion: v1
kind: Secret
metadata:
  labels:
    app: release-name-mysql
    chart: mysql-1.3.2
    heritage: Tiller
    release: release-name
  name: release-name-mysql
  namespace: default
stringData:
  mysql-password: refs+vault://secret/data/foo#/mykey
  mysql-root-password: vZQmqdGw3z
type: Opaque

This manifest is safe to be committed into your version-control system(GitOps!) as it doesn't contain actual secrets.

When you finally deploy the manifests, run vals eval to replace all the refs to actual secrets:

$ cat manifests.yaml | ~/p/values/bin/vals eval -f - | tee all.yaml
apiVersion: v1
kind: Secret
metadata:
    labels:
        app: release-name-mysql
        chart: mysql-1.3.2
        heritage: Tiller
        release: release-name
    name: release-name-mysql
    namespace: default
stringData:
    mysql-password: myvalue
    mysql-root-password: 0A8V1SER9t
type: Opaque

Finally run kubectl apply to apply manifests:

$ kubectl apply -f all.yaml

This gives you a solid foundation for building a secure CD system as you need to allow access to a secrets store like Vault only from servers or containers that pulls safe manifests and runs deployments.

In other words, you can safely omit access from the CI to the secrets store.

Go

import "github.com/variantdev/vals"

secretsToCache := 256 // how many secrets to keep in LRU cache
runtime, err := vals.New(secretsToCache)
if err != nil {
  return nil, err
}

valsRendered, err := runtime.Eval(map[string]interface{}{
    "inline": map[string]interface{}{
        "foo": "ref+vault://127.0.0.1:8200/mykv/foo?proto=http#/mykey",
        "bar": map[string]interface{}{
            "baz": "ref+vault://127.0.0.1:8200/mykv/foo?proto=http#/mykey",
        },
    },
})

Now, vals contains a map[string]interface{} representation of the below:

cat <<EOF
foo: $(vault read mykv/foo -o json | jq -r .mykey)
  bar:
    baz: $(vault read mykv/foo -o json | jq -r .mykey)
EOF

Supported Backends

• Vault
• AWS SSM Parameter Store
• AWS Secrets Manager
• AWS S3
• GCP Secrets Manager
• Google GCS
• SOPS powered by sops
• Terraform (tfstate) powered by tfstate-lookup
• Echo
• File
• Azure Key Vault

Please see pkg/providers for the implementations of all the providers. The package names corresponds to the URI schemes.

Vault

• ref+vault://PATH/TO/KVBACKEND[?address=VAULT_ADDR:PORT&token_file=PATH/TO/FILE&token_env=VAULT_TOKEN&namespace=VAULT_NAMESPACE]#/fieldkey
• ref+vault://PATH/TO/KVBACKEND[?address=VAULT_ADDR:PORT&auth_method=approle&role_id=ce5e571a-f7d4-4c73-93dd-fd6922119839&secret_id=5c9194b9-585e-4539-a865-f45604bd6f56]#/fieldkey
• ref+vault://PATH/TO/KVBACKEND[?address=VAULT_ADDR:PORT&auth_method=kubernetes&role_id=K8S-ROLE

• address defaults to the value of the VAULT_ADDR envvar.
• namespace defaults to the value of the VAULT_NAMESPACE envvar.
• auth_method default to token and can also be set to the value of the VAULT_AUTH_METHOD envar.
• role_id defaults to the value of the VAULT_ROLE_ID envvar.
• secret_id defaults to the value of the VAULT_SECRET_ID envvar.
• version is the specific version of the secret to be obtained. Used when you want to get a previous content of the secret.

Authentication

The auth_method or VAULT_AUTH_METHOD envar configures how vals authenticates to HashiCorp Vault. Currently only these options are supported:

• approle: it requires you pass on a role_id together with a secret_id.
• token: you just need creating and passing on a VAULT_TOKEN
• kubernetes: if you're running inside a Kubernetes cluster, you can use this option. It requires you configure a policy, a Kubernetes role, a service account and a JWT token. The login path can also be set using the environment variable VAULT_KUBERNETES_MOUNT_POINT (default is /kubernetes). You must also set role_id or VAULT_ROLE_ID envar to the Kubernetes role.

Examples:

• ref+vault://mykv/foo#/bar?address=https://vault1.example.com:8200 reads the value for the field bar in the kv foo on Vault listening on https://vault1.example.com with the Vault token read from the envvar VAULT_TOKEN, or the file ~/.vault_token when the envvar is not set
• ref+vault://mykv/foo#/bar?token_env=VAULT_TOKEN_VAULT1&namespace=ns1&address=https://vault1.example.com:8200 reads the value for the field bar from namespace ns1 in the kv foo on Vault listening on https://vault1.example.com with the Vault token read from the envvar VAULT_TOKEN_VAULT1
• ref+vault://mykv/foo#/bar?token_file=~/.vault_token_vault1&address=https://vault1.example.com:8200 reads the value for the field bar in the kv foo on Vault listening on https://vault1.example.com with the Vault token read from the file ~/.vault_token_vault1
• ref+vault://mykv/foo#/bar?role_id=my-kube-role using the Kubernetes role to log in to Vault

AWS

There are two providers for AWS:

• SSM Parameter Store
• Secrets Manager

Both provider have support for specifying AWS region and profile via envvars or options:

• AWS profile can be specified via an option profile=AWS_PROFILE_NAME or envvar AWS_PROFILE
• AWS region can be specified via an option region=AWS_REGION_NAME or envvar AWS_DEFAULT_REGION

AWS SSM Parameter Store

• ref+awsssm://PATH/TO/PARAM[?region=REGION]
• ref+awsssm://PREFIX/TO/PARAMS[?region=REGION&mode=MODE&version=VERSION]#/PATH/TO/PARAM

The first form result in a GetParameter call and result in the reference to be replaced with the value of the parameter.

The second form is handy but fairly complex.

• If mode is not set, vals uses GetParametersByPath(/PREFIX/TO/PARAMS) caches the result per prefix rather than each single path to reduce number of API calls
• If mode is singleparam, vals uses GetParameter to obtain the value parameter for key /PREFIX/TO/PARAMS, parse the value as a YAML hash, extract the value at the yaml path PATH.TO.PARAM.
  • When version is set, vals uses GetParameterHistoryPages instead of GetParameter.

For the second form, you can optionally specify recursive=true to enable the recursive option of the GetParametersByPath API.

Let's say you had a number of parameters like:

NAME        VALUE
/foo/bar    {"BAR":"VALUE"}
/foo/bar/a  A
/foo/bar/b  B

• ref+awsssm://foo/bar and ref+awsssm://foo#/bar results in {"BAR":"VALUE"}
• ref+awsssm://foo/bar/a, ref+awsssm://foo/bar?#/a, and ref+awsssm://foo?recursive=true#/bar/a results in A
• ref+awsssm://foo/bar?mode=singleparam#/BAR results in VALUE.

On the other hand,

• ref+awsssm://foo/bar#/BAR fails because /foo/bar evaluates to {"a":"A","b":"B"}.
• ref+awsssm://foo?recursive=true#/bar fails because /foo?recursive=true internal evaluates to {"foo":{"a":"A","b":"B"}}

AWS Secrets Manager

• ref+awssecrets://PATH/TO/SECRET[?region=REGION&version_stage=STAGE&version_id=ID]
• ref+awssecrets://PATH/TO/SECRET[?region=REGION&version_stage=STAGE&version_id=ID]#/yaml_or_json_key/in/secret
• ref+awssecrets://ACCOUNT:ARN:secret:/PATH/TO/PARAM[?region=REGION]

The third form allows you to reference a secret in another AWS account (if your cross-account secret permissions are configured).

Examples:

• ref+awssecrets://myteam/mykey
• ref+awssecrets://myteam/mydoc#/foo/bar
• ref+awssecrets://myteam/mykey?region=us-west-2
• ref+awssecrets:///arn:aws:secretsmanager:<REGION>:<ACCOUNT_ID>:secret:/myteam/mydoc/?region=ap-southeast-2#/secret/key

AWS S3

• ref+s3://BUCKET/KEY/OF/OBJECT[?region=REGION&profile=AWS_PROFILE&version_id=ID]
• ref+s3://BUCKET/KEY/OF/OBJECT[?region=REGION&profile=AWS_PROFILE&version_id=ID]#/yaml_or_json_key/in/secret

Examples:

• ref+s3://mybucket/mykey
• ref+s3://mybucket/myjsonobj#/foo/bar
• ref+s3://mybucket/myyamlobj#/foo/bar
• ref+s3://mybucket/mykey?region=us-west-2
• ref+s3://mybucket/mykey?profile=prod

Google GCS

• ref+gcs://BUCKET/KEY/OF/OBJECT[?generation=ID]
• ref+gcs://BUCKET/KEY/OF/OBJECT[?generation=ID]#/yaml_or_json_key/in/secret

Examples:

• ref+gcs://mybucket/mykey
• ref+gcs://mybucket/myjsonobj#/foo/bar
• ref+gcs://mybucket/myyamlobj#/foo/bar
• ref+gcs://mybucket/mykey?generation=1639567476974625

GCP Secrets Manager

• ref+gcpsecrets://PROJECT/SECRET[?version=VERSION]
• ref+gcpsecrets://PROJECT/SECRET[?version=VERSION]#/yaml_or_json_key/in/secret

Examples:

• ref+gcpsecrets://myproject/mysecret
• ref+gcpsecrets://myproject/mysecret?version=3
• ref+gcpsecrets://myproject/mysecret?version=3#/yaml_or_json_key/in/secret

NOTE: Got an error like expand gcpsecrets://project/secret-name?version=1: failed to get secret: rpc error: code = PermissionDenied desc = Request had insufficient authentication scopes.?

In some cases like you need to use an alternative credentials or project, you'll likely need to set GOOGLE_APPLICATION_CREDENTIALS and/or GCP_PROJECT envvars.

Terraform (tfstate)

• ref+tfstate://relative/path/to/some.tfstate/RESOURCE_NAME
• ref+tfstate:///absolute/path/to/some.tfstate/RESOURCE_NAME

Examples:

• ref+tfstate://path/to/some.tfstate/aws_vpc.main.id
• ref+tfstate://path/to/some.tfstate/module.mymodule.aws_vpc.main.id
• ref+tfstate://path/to/some.tfstate/output.OUTPUT_NAME
• ref+tfstate://path/to/some.tfstate/data.thetype.name.foo.bar

When you're using terraform-aws-vpc to define a module "vpc" resource and you wanted to grab the first vpc ARN created by the module:

$ tfstate-lookup -s ./terraform.tfstate module.vpc.aws_vpc.this[0].arn
arn:aws:ec2:us-east-2:ACCOUNT_ID:vpc/vpc-0cb48a12e4df7ad4c

$ echo 'foo: ref+tfstate://terraform.tfstate/module.vpc.aws_vpc.this[0].arn' | vals eval -f -
foo: arn:aws:ec2:us-east-2:ACCOUNT_ID:vpc/vpc-0cb48a12e4df7ad4c

You can also grab a Terraform output by using output.OUTPUT_NAME like:

$ tfstate-lookup -s ./terraform.tfstate output.mystack_apply

which is equivalent to the following input for vals:

$ echo 'foo: ref+tfstate://terraform.tfstate/output.mystack_apply' | vals eval -f -

Remote backends like S3, GCS and AzureRM blob store are also supported. When a remote backend is used in your terraform workspace, there should be a local file at ./terraform/terraform.tfstate that contains the reference to the backend:

{
    "version": 3,
    "serial": 1,
    "lineage": "f1ad69de-68b8-9fe5-7e87-0cb70d8572c8",
    "backend": {
        "type": "s3",
        "config": {
            "access_key": null,
            "acl": null,
            "assume_role_policy": null,
            "bucket": "yourbucketnname",

Just specify the path to that file, so that vals is able to transparently make the remote state contents available for you.

Terraform in GCS bucket (tfstategs)

• ref+tfstategs://bucket/path/to/some.tfstate/RESOURCE_NAME

Examples:

• ref+tfstategs://bucket/path/to/some.tfstate/google_compute_disk.instance.id

It allows to use Terraform state stored in GCS bucket with the direct URL to it. You can try to read the state with command:

$ tfstate-lookup -s gs://bucket-with-terraform-state/terraform.tfstate google_compute_disk.instance.source_image_id
5449927740744213880

which is equivalent to the following input for vals:

$ echo 'foo: ref+tfstategs://bucket-with-terraform-state/terraform.tfstate/google_compute_disk.instance.source_image_id' | vals eval -f -

Terraform in S3 bucket (tfstates3)

• ref+tfstates3://bucket/path/to/some.tfstate/RESOURCE_NAME

Examples:

• ref+tfstates3://bucket/path/to/some.tfstate/aws_vpc.main.id

It allows to use Terraform state stored in AWS S3 bucket with the direct URL to it. You can try to read the state with command:

$ tfstate-lookup -s s3://bucket-with-terraform-state/terraform.tfstate module.vpc.aws_vpc.this[0].arn
arn:aws:ec2:us-east-2:ACCOUNT_ID:vpc/vpc-0cb48a12e4df7ad4c

which is equivalent to the following input for vals:

$ echo 'foo: ref+tfstates3://bucket-with-terraform-state/terraform.tfstate/module.vpc.aws_vpc.this[0].arn' | vals eval -f -

Terraform in AzureRM Blob storage (tfstateazurerm)

• ref+tfstateazurerm://{resource_group_name}/{storage_account_name}/{container_name}/{blob_name}.tfstate/RESOURCE_NAME

Examples:

• ref+tfstateazurerm://my_rg/my_storage_account/terraform-backend/unique.terraform.tfstate/output.virtual_network.name

It allows to use Terraform state stored in Azure Blob storage given the resource group, storage account, container name and blob name. You can try to read the state with command:

$ tfstate-lookup -s azurerm://my_rg/my_storage_account/terraform-backend/unique.terraform.tfstate output.virtual_network.name

which is equivalent to the following input for vals:

$ echo 'foo: ref+tfstateazurerm://my_rg/my_storage_account/terraform-backend/unique.terraform.tfstate/output.virtual_network.name' | vals eval -f -

Terraform in Terraform Cloud / Terraform Enterprise (tfstateremote)

• ref+tfstateremote://app.terraform.io/{org}/{myworkspace}/RESOURCE_NAME

Examples:

• ref+tfstateremote://app.terraform.io/myorg/myworkspace/output.virtual_network.name

It allows to use Terraform state stored in Terraform Cloud / Terraform Enterprise given the resource group, the organization and the workspace. You can try to read the state with command (with exported variable TFE_TOKEN):

$ tfstate-lookup -s remote://app.terraform.io/myorg/myworkspace output.virtual_network.name

which is equivalent to the following input for vals:

$ echo 'foo: ref+tfstateremote://app.terraform.io/myorg/myworkspace/output.virtual_network.name' | vals eval -f -

SOPS

• The whole content of a SOPS-encrypted file: ref+sops://base64_data_or_path_to_file?key_type=[filepath|base64]&format=[binary|dotenv|yaml]
• The value for the specific path in an encrypted YAML/JSON document: ref+sops://base64_data_or_path_to_file#/json_or_yaml_key/in/the_encrypted_doc

Examples:

• ref+sops://path/to/file reads path/to/file as binary input
• ref+sops://<base64>?key_type=base64 reads <base64> as the base64-encoded data to be decrypted by sops as binary
• ref+sops://path/to/file#/foo/bar reads path/to/file as a yaml file and returns the value at foo.bar.
• ref+sops://path/to/file?format=json#/foo/bar reads path/to/file as a json file and returns the value at foo.bar.

Echo

Echo provider echoes the string for testing purpose. Please read the original proposal to get why we might need this.

• ref+echo://KEY1/KEY2/VALUE[#/path/to/the/value]

Examples:

• ref+echo://foo/bar generates foo/bar
• ref+echo://foo/bar/baz#/foo/bar generates baz. This works by the host and the path part foo/bar/baz generating an object {"foo":{"bar":"baz"}} and the fragment part #/foo/bar results in digging the object to obtain the value at $.foo.bar.

File

File provider reads a local text file, or the value for the specific path in a YAML/JSON file.

• ref+file://relative/path/to/file[#/path/to/the/value]
• ref+file:///absolute/path/to/file[#/path/to/the/value]

Examples:

• ref+file://foo/bar loads the file at foo/bar
• ref+file:///home/foo/bar loads the file at /home/foo/bar
• ref+file://some.yaml#/foo/bar loads the YAML file at some.yaml and reads the value for the path $.foo.bar. Let's say some.yaml contains {"foo":{"bar":"BAR"}}, key1: ref+file://some.yaml#/foo/bar results in key1: BAR.

Azure Key Vault

Retrieve secrets from Azure Key Vault. Path is used to specify the vault and secret name. Optionally a specific secret version can be retrieved.

• ref+azurekeyvault://VAULT-NAME/SECRET-NAME[/VERSION]

VAULT-NAME is either a simple name if operating in AzureCloud (vault.azure.net) or the full endpoint dns name when operating against non-default azure clouds (US Gov Cloud, China Cloud, German Cloud). Examples:

• ref+azurekeyvault://my-vault/secret-a
• ref+azurekeyvault://my-vault/secret-a/ba4f196b15f644cd9e949896a21bab0d
• ref+azurekeyvault://gov-cloud-test.vault.usgovcloudapi.net/secret-b

Authentication

Vals aquires Azure credentials though Azure CLI or from environment variables. The easiest way is to run az login. Vals can then aquire the current credentials from az without further set up.

Other authentication methods require information to be passed in environment variables. See Azure SDK docs and auth.go for the full list of supported environment variables.

For example, if using client credentials the required env vars are AZURE_CLIENT_ID, AZURE_CLIENT_SECRET, AZURE_TENANT_ID and possibly AZURE_ENVIRONMENT in case of accessing an Azure GovCloud.

The order in which authentication methods are checked is:

1. Client credentials
2. Client certificate
3. Username/Password
4. Azure CLI or Managed identity (set environment AZURE_USE_MSI=true to enabled MSI)

Advanced Usages

Discriminating config and secrets

vals has an advanced feature that helps you to do GitOps.

GitOps is a good practice that helps you to review how your change would affect the production environment.

To best leverage GitOps, it is important to remove dynamic aspects of your config before reviewing.

On the other hand, vals's primary purpose is to defer retrieval of values until the time of deployment, so that we won't accidentally git-commit secrets. The flip-side of this is, obviously, that you can't review the values themselves.

Using ref+<value uri> and secretref+<value uri> in combination with vals eval --exclude-secretref helps it.

By using the secretref+<uri> notation, you tell vals that it is a secret and regular ref+<uri> instances are for config values.

myconfigvalue: ref+awsssm://myconfig/value
mysecretvalue: secretref+awssecrets://mysecret/value

To leverage GitOps most by allowing you to review the content of ref+awsssm://myconfig/value only, you run vals eval --exclude-secretref to generate the following:

myconfigvalue: MYCONFIG_VALUE
mysecretvalue: secretref+awssecrets://mysecret/value

This is safe to be committed into git because, as you've told to vals, awsssm://myconfig/value is a config value that can be shared publicly.

Non-Goals

String-Interpolation / Template Functions

In the early days of this project, the original author has investigated if it was a good idea to introduce string interpolation like feature to vals:

foo: xx${{ref "ref+vault://127.0.0.1:8200/mykv/foo?proto=http#/mykey" }}
bar:
  baz: yy${{ref "ref+vault://127.0.0.1:8200/mykv/foo?proto=http#/mykey" }}

But the idea had abandoned due to that it seemed to drive the momentum to vals being a full-fledged YAML templating engine. What if some users started wanting to use vals for transforming values with functions? That's not the business of vals.

Instead, use vals solely for composing sets of values that are then input to another templating engine or data manipulation language like Jsonnet and CUE.

Merge

Merging YAMLs is out of the scope of vals. There're better alternatives like Jsonnet, Sprig, and CUE for the job.