Berglas is a command line tool and library for storing and retrieving secrets on Google Cloud. Secrets are encrypted with Cloud KMS and stored in Cloud Storage. An interoperable layer also exists with Secret Manager.
-
As a CLI,
berglas
automates the process of encrypting, decrypting, and storing data on Google Cloud. -
As a library,
berglas
automates the inclusion of secrets into various Google Cloud runtimes
Berglas is not an officially supported Google product.
-
Install the Cloud SDK for your operating system. Alternatively, you can run these commands from Cloud Shell, which has the SDK and other popular tools pre-installed.
If you are running from your local machine, you also need Default Application Credentials:
gcloud auth application-default login
This will open a web browser and prompt for a login to your Google account. On headless devices, you will need to create a service account. For more information, please see the authentication section.
-
Install the
berglas
CLI using one of the following methods:-
Install a pre-compiled binary for your operating system:
This will download the latest version built against the master branch. To download a specific version, specify a git tag in place of "master" in the URL.
Depending on your operating system, you may need to mark the downloaded binary as executable:
chmod +x /path/to/berglas
-
If you use Homebrew on macOS, you can install like this:
brew install berglas
-
Use the official Docker container:
docker pull gcr.io/berglas/berglas:latest
This will pull the latest version built against the master branch. To use a specific version, specify a git tag in place of "latest" in the URL.
-
Install from source (requires a working Go installation):
go get github.com/GoogleCloudPlatform/berglas/... go install github.com/GoogleCloudPlatform/berglas
-
-
Export your project ID as an environment variable. The rest of this setup guide assumes this environment variable is set:
export PROJECT_ID=my-gcp-project-id
Please note, this is the project ID, not the project name or project number. You can find the project ID by running
gcloud projects list
or in the web UI. -
Export your desired Cloud Storage bucket name. The rest of this setup guide assumes this environment variable is set:
export BUCKET_ID=my-secrets
Replace
my-secrets
with the name of your bucket. Set only the name, without thegs://
prefix. This bucket should not exist yet! -
Enable required services on the project:
gcloud services enable --project ${PROJECT_ID} \ cloudkms.googleapis.com \ storage-api.googleapis.com \ storage-component.googleapis.com
-
Bootstrap a Berglas environment. This will create a new Cloud Storage bucket for storing secrets and a Cloud KMS key for encrypting data.
berglas bootstrap --project $PROJECT_ID --bucket $BUCKET_ID
This command uses the default values. You can customize the storage bucket and KMS key configuration using the optional flags. Run
berglas bootstrap -h
for more details.If you want full control over the creation of the Cloud Storage and Cloud KMS keys, please see the custom setup documentation.
-
(Optional) Bootstrap a Berglas environment specifying a bucket location. By default the berglas bucket is created in the multi-regional location
US
. You can specify your location by using the following command. Please see the list of supported locations in the GCP bucket location documentation pageexport BUCKET_LOCATION=europe-west1 berglas bootstrap \ --project $PROJECT_ID \ --bucket $BUCKET_ID \ --bucket-location $BUCKET_LOCATION
This command uses the default values. You can customize the storage bucket and KMS key configuration using the optional flags. Run
berglas bootstrap -h
for more details.If you want full control over the creation of the Cloud Storage and Cloud KMS keys, please see the custom setup documentation.
-
(Optional) Enable Cloud Audit logging on the bucket:
Please note this will enable audit logging on all Cloud KMS keys and all Cloud Storage buckets in the project, which may incur additional costs.
-
Download the exiting project IAM policy:
gcloud projects get-iam-policy ${PROJECT_ID} > policy.yaml
-
Add Cloud Audit logging for Cloud KMS and Cloud Storage:
cat <<EOF >> policy.yaml auditConfigs: - auditLogConfigs: - logType: DATA_READ - logType: ADMIN_READ - logType: DATA_WRITE service: cloudkms.googleapis.com - auditLogConfigs: - logType: ADMIN_READ - logType: DATA_READ - logType: DATA_WRITE service: storage.googleapis.com EOF
-
Submit the new policy:
gcloud projects set-iam-policy ${PROJECT_ID} policy.yaml
-
Remove the updated policy from local disk:
rm policy.yaml
-
-
Create a secret:
berglas create ${BUCKET_ID}/foo my-secret-data \ --key projects/${PROJECT_ID}/locations/global/keyRings/berglas/cryptoKeys/berglas-key Successfully created secret: foo
-
Grant access to a secret:
berglas grant ${BUCKET_ID}/foo --member user:[email protected]
-
Access a secret's data:
berglas access ${BUCKET_ID}/foo my-secret-data
-
Spawn a child process with secrets populated in the child's environment:
berglas exec -- myapp --flag-a --flag-b
This will spawn
myapp
with an environment parsed by berglas. -
Access data from a specific generation of a secret:
berglas access ${BUCKET_ID}/foo#1563925940580201 my-previous-secret-data
-
Revoke access to a secret:
berglas revoke ${BUCKET_ID}/foo --member user:[email protected]
-
Delete a secret:
berglas delete ${BUCKET_ID}/foo Successfully deleted secret if it existed: foo
In addition to standard Unix exit codes, if the CLI exits with a known error, Berglas will exit with one of the following:
-
60
- API error. Berglas got a bad response when communicating with an upstream API. -
61
- Misuse error. You gave unexpected input or behavior. Please read the error message. Open an issue if you think this is a mistake.
The only exception is berglas exec
, which will exit with the exit status of
its child command, if one was provided.
-
App Engine (Flex) - When invoked via App Engine Flex, Berglas resolves environment variables to their plaintext values using the `berglas://reference syntax. This integration works with any language runtime because berglas serves as the entrypoint to the Docker container. See examples/appengineflex for examples and invocations.
-
App Engine (Standard) - When invoked via App Engine, Berglas resolves environment variables to their plaintext values using the
berglas://
reference syntax. This integration only works with the Go language runtime because it requires importing theauto/
package. See examples/appengine for examples and invocations. -
Cloud Run - When invoked via Cloud Run, Berglas resolves environment variables to their plaintext values using the
berglas://
reference syntax. This integration works with any language runtime because berglas serves as the entrypoint to the Docker container. See examples/cloudrun for examples and invocations. -
Cloud Functions - When invoked via Cloud Functions, Berglas resolves environment variables to their plaintext values using the
berglas://
reference syntax. This integration only works with the Go language runtime because it requires importing theauto/
package. See examples/cloudfunctions for examples and invocations. -
Cloud Build - When invoked via Cloud Build, Berglas resolves environment variables to plaintext values using the
berglas://
reference syntax. This integration only works with volume mounts, so all Berglas secrets need to specify the?destination
parameter. See examples/cloudbuild for examples and invocations. -
Kubernetes - Kubernetes pods can consume Berglas secrets by installing a MutatingWebhook. This webhook mutates incoming pods with the
berglas://
reference syntax in environment references to resolve at runtime. This integration works with any container, but all pods requesting berglas secrets must set an command in their Kubernetes manifests. See examples/kubernetes for samples and installation instructions. -
Anything - Wrap any process with
berglas exec --
and Berglas will parse any local environment variables with theberglas://
reference syntax and spawn your app as a subprocess with the plaintext environment replaced.
Both the berglas CLI and berglas library support debug-style logging. This logging is off by default because it adds additional overhead and logs information that may be security-sensitive.
The default logging behavior for the berglas CLI is "text" (it can be changed
with the --log-format
flag). The default logging behavior for the berglas
library is structured JSON which integrates well with Stackdriver (it can be
changed to any valid formatter and you can even inject your own logger).
Examples are available in the examples/
folder.
Berglas is also a Go library that can be imported in Go projects:
import (
_ "github.com/GoogleCloudPlatform/berglas/pkg/auto"
)
When imported, the berglas
package will:
-
Download and decrypt any secrets that match the Berglas environment variable reference syntax in the environment
-
Replace the value for the environment variable with the decrypted secret
You can also opt out of auto-parsing and call the library yourself instead:
import (
"context"
"log"
"os"
"github.com/GoogleCloudPlatform/berglas/pkg/berglas"
)
func main() {
ctx := context.Background()
// This higher-level API parses the secret reference at the specified
// environment variable, downloads and decrypts the secret, and replaces the
// contents of the given environment variable with the secret result.
if err := berglas.Replace(ctx, "MY_SECRET"); err != nil {
log.Fatal(err)
}
// This lower-level API parses the secret reference, downloads and decrypts
// the secret, and returns the result. This is useful if you need to mutate
// the result.
if v := os.Getenv("MY_SECRET"); v != "" {
plaintext, err := berglas.Resolve(ctx, v)
if err != nil {
log.Fatal(err)
}
os.Unsetenv("MY_SECRET")
os.Setenv("MY_OTHER_SECRET", string(plaintext))
}
}
For more examples and documentation, please see the godoc.
By default, Berglas uses Google Cloud Default Application Credentials. If you have gcloud installed locally, ensure you have application default credentials:
gcloud auth application-default login
On GCP services (like Cloud Build, Compute, etc), it will use the service account attached to the resource.
To use a specific service account, set the GOOGLE_APPLICATION_CREDENTIALS
environment variable to the filepath to the JSON file where your credentials
reside on disk:
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/my/credentials.json
To learn more, please see the Google Cloud Service Account documentation.
To control who or what has access to a secret, use berglas grant
and berglas revoke
or the associated API methods. These methods use Cloud IAM
internally. The following information is relevant only if you plan to grant IAM
permissions manually.
Most operations require access to the Cloud KMS key and the Cloud Storage bucket. You can read more about Cloud KMS IAM and Cloud Storage IAM in the documentation.
Any service account or entity using Berglas will need to authorize using the
cloud-platform
scope.
To create a secret, the following role is required on the Cloud Storage bucket:
roles/storage.objectCreator
To create a secret, the following role is required for the Cloud KMS key:
roles/cloudkms.cryptoKeyEncrypter
To access a secret, the following role is required on the Cloud Storage bucket:
roles/storage.objectViewer
To access a secret, the following role is required for the Cloud KMS key:
roles/cloudkms.cryptoKeyDecrypter
To delete a secret, the following role is required on the Cloud Storage bucket:
roles/storage.objectAdmin
To delete a secret, no permissions are needed on the Cloud KMS key.
This section describes the implementation. This knowledge is not required to use Berglas, but it is included for security-conscious/curious users who want to learn about how Berglas works internally to build a threat model.
When encrypting a secret:
-
Berglas generates an AES-256-GCM data encryption key (DEK) using Go's crypto package for each secret. (N.B. each secret has its own, unique DEK).
-
Berglas encrypts the plaintext data using the locally-generated DEK, producing encrypted ciphertext, prepended with the AES-GCM nonce.
-
Berglas encrypts the DEK using the specified Cloud KMS key, also known as a key encryption key (KEK). This process is called envelope encryption.
-
Berglas stores the Cloud KMS key name, encrypted DEK, and encrypted ciphertext as a single blob in Cloud Storage.
When decrypting a secret:
-
Berglas downloads the blob from Cloud Storage and separates the Cloud KMS key name, encrypted DEK, and ciphertext out of the blob.
-
Berglas decrypts the DEK using Cloud KMS. This is part of envelope encryption.
-
Berglas decrypts the ciphertext data locally using the decrypted DEK.
Berglas makes certain security tradeoffs in exchange for a better UX. In particular, KMS crypto key IDs are stored on the secret object's metadata. An attacker with permission to write objects to your Cloud Storage bucket could overwrite existing secrets. As such, you should follow the principles of least privilege and revoke default ACLs on the bucket as described in the setup guide.
For more information, please see the security and threat model.
Q: Should I use Berglas or Secret Manager?
Berglas is compatible with Secret Manager and offers
convenience wrappers around managing secrets regardless of whether they reside
in Cloud Storage or Secret Manager. New projects should investigate using Secret
Manager directly as it has less operational overhead and complexity, but Berglas
will continue to support Cloud Storage + Cloud KMS secrets.
Q: Is there a size limit on the data I can encrypt?
Berglas is targeted at application secrets like certificates, passwords, and
API keys. While its possible to encrypt larger binary files like PDFs or images,
Berglas uses a a GCM cipher mode to encrypt data, meaning the data must fit in
memory and is limited to 64GiB.
Q: Why do you use envelope encryption instead of
encrypting the data directly with Cloud KMS?
Envelope encryption allows for encrypting the data at the application layer,
and it enables encryption of larger payloads, since Cloud KMS has a limit on the
size of the payload it can encrypt. By using envelope encryption, Cloud KMS
always encrypts a fixed size data (the AES-256-GCM key). This saves bandwidth
(since large payloads are encrypted locally) and increases the size of the data
which can be encrypted.
Q: Why does Berglas need permission to view my GCP resource?
Berglas communicates with the API to read the environment variables that were
set on the resource at deploy time. Otherwise, a package could inject arbitrary
environment variables in the Berglas format during application boot.
Q: I renamed a secret in Cloud Storage and now it fails to decrypt - why?
Berglas encrypts secrets with additional authenticated data including the name
of the secret. This reduces the chance an attacker can escalate privilege by
convincing someone to rename a secret so they can gain access.
Q: Why is it named Berglas?
Berglas is a famous magician who is best known for his secrets.
Please see the contributing guidelines.
This library is licensed under Apache 2.0. Full license text is available in LICENSE.