Add support for reading private_key from AWS Secrets Manager (#31134)

Adds support for reading server `private_key` from AWS Secrets Manager.
Combined with #31075, this should allow removing all common sensitive
secrets from the environment/config (if I missed any let me know). This
works with localstack for local development (set
`AWS_ENDPOINT_URL=$LOCALSTACK_URL`, `AWS_ACCESS_KEY_ID=test`, and
`AWS_SECRET_ACCESS_KEY=test`).

I did not include config options for `AWS_ACCESS_KEY_ID` and
`AWS_SECRET_ACCESS_KEY` because they are a bad practice vs role
credentials and defeat the purpose of this feature which is to remove
secrets from the environment/config.

# Checklist for submitter

If some of the following don't apply, delete the relevant line.

<!-- Note that API documentation changes are now addressed by the
product design team. -->

- [x] Changes file added for user-visible changes in `changes/`,
`orbit/changes/` or `ee/fleetd-chrome/changes`.
See [Changes
files](https://github.com/fleetdm/fleet/blob/main/docs/Contributing/guides/committing-changes.md#changes-files)
for more information.
- [x] Added/updated automated tests
- [x] Manual QA for all new/changed functionality

---------

Co-authored-by: Scott Gress <scott@fleetdm.com>
This commit is contained in:
Jonathan Rudenberg
2025-09-09 16:56:35 -05:00
committed by GitHub
co-authored by Scott Gress
parent 09c6561f65
commit 48760fec58
8 changed files with 471 additions and 38 deletions
+37 -28
View File
@@ -95,20 +95,23 @@ const (
// ServerConfig defines configs related to the Fleet server
type ServerConfig struct {
Address string
Cert string
Key string
TLS bool
TLSProfile string `yaml:"tls_compatibility"`
URLPrefix string `yaml:"url_prefix"`
Keepalive bool `yaml:"keepalive"`
SandboxEnabled bool `yaml:"sandbox_enabled"`
WebsocketsAllowUnsafeOrigin bool `yaml:"websockets_allow_unsafe_origin"`
FrequentCleanupsEnabled bool `yaml:"frequent_cleanups_enabled"`
ForceH2C bool `yaml:"force_h2c"`
PrivateKey string `yaml:"private_key"`
VPPVerifyTimeout time.Duration `yaml:"vpp_verify_timeout"`
VPPVerifyRequestDelay time.Duration `yaml:"vpp_verify_request_delay"`
Address string
Cert string
Key string
TLS bool
TLSProfile string `yaml:"tls_compatibility"`
URLPrefix string `yaml:"url_prefix"`
Keepalive bool `yaml:"keepalive"`
SandboxEnabled bool `yaml:"sandbox_enabled"`
WebsocketsAllowUnsafeOrigin bool `yaml:"websockets_allow_unsafe_origin"`
FrequentCleanupsEnabled bool `yaml:"frequent_cleanups_enabled"`
ForceH2C bool `yaml:"force_h2c"`
PrivateKey string `yaml:"private_key"`
PrivateKeySecretArn string `yaml:"private_key_secret_arn"`
PrivateKeySecretSTSAssumeRoleArn string `yaml:"private_key_secret_sts_assume_role_arn"`
PrivateKeySecretSTSExternalID string `yaml:"private_key_secret_sts_external_id"`
VPPVerifyTimeout time.Duration `yaml:"vpp_verify_timeout"`
VPPVerifyRequestDelay time.Duration `yaml:"vpp_verify_request_delay"`
}
func (s *ServerConfig) DefaultHTTPServer(ctx context.Context, handler http.Handler) *http.Server {
@@ -1121,6 +1124,9 @@ func (man Manager) addConfigs() {
man.addConfigBool("server.frequent_cleanups_enabled", false, "Enable frequent cleanups of expired data (15 minute interval)")
man.addConfigBool("server.force_h2c", false, "Force the fleet server to use HTTP2 cleartext aka h2c (ignored if using TLS)")
man.addConfigString("server.private_key", "", "Used for encrypting sensitive data, such as MDM certificates.")
man.addConfigString("server.private_key_secret_arn", "", "ARN of AWS Secrets Manager secret containing server private key")
man.addConfigString("server.private_key_secret_sts_assume_role_arn", "", "ARN of role to assume for accessing private key secret")
man.addConfigString("server.private_key_secret_sts_external_id", "", "External ID for STS role assumption when accessing private key secret")
man.addConfigDuration("server.vpp_verify_timeout", 10*time.Minute, "Maximum amout of time to wait for VPP app install verification")
man.addConfigDuration("server.vpp_verify_request_delay", 5*time.Second, "Delay in between requests to verify VPP app installs")
@@ -1548,20 +1554,23 @@ func (man Manager) LoadConfig() FleetConfig {
StsExternalID: man.getConfigString("redis.sts_external_id"),
},
Server: ServerConfig{
Address: man.getConfigString("server.address"),
Cert: man.getConfigString("server.cert"),
Key: man.getConfigString("server.key"),
TLS: man.getConfigBool("server.tls"),
TLSProfile: man.getConfigTLSProfile(),
URLPrefix: man.getConfigString("server.url_prefix"),
Keepalive: man.getConfigBool("server.keepalive"),
SandboxEnabled: man.getConfigBool("server.sandbox_enabled"),
WebsocketsAllowUnsafeOrigin: man.getConfigBool("server.websockets_allow_unsafe_origin"),
FrequentCleanupsEnabled: man.getConfigBool("server.frequent_cleanups_enabled"),
ForceH2C: man.getConfigBool("server.force_h2c"),
PrivateKey: man.getConfigString("server.private_key"),
VPPVerifyTimeout: man.getConfigDuration("server.vpp_verify_timeout"),
VPPVerifyRequestDelay: man.getConfigDuration("server.vpp_verify_request_delay"),
Address: man.getConfigString("server.address"),
Cert: man.getConfigString("server.cert"),
Key: man.getConfigString("server.key"),
TLS: man.getConfigBool("server.tls"),
TLSProfile: man.getConfigTLSProfile(),
URLPrefix: man.getConfigString("server.url_prefix"),
Keepalive: man.getConfigBool("server.keepalive"),
SandboxEnabled: man.getConfigBool("server.sandbox_enabled"),
WebsocketsAllowUnsafeOrigin: man.getConfigBool("server.websockets_allow_unsafe_origin"),
FrequentCleanupsEnabled: man.getConfigBool("server.frequent_cleanups_enabled"),
ForceH2C: man.getConfigBool("server.force_h2c"),
PrivateKey: man.getConfigString("server.private_key"),
PrivateKeySecretArn: man.getConfigString("server.private_key_secret_arn"),
PrivateKeySecretSTSAssumeRoleArn: man.getConfigString("server.private_key_secret_sts_assume_role_arn"),
PrivateKeySecretSTSExternalID: man.getConfigString("server.private_key_secret_sts_external_id"),
VPPVerifyTimeout: man.getConfigDuration("server.vpp_verify_timeout"),
VPPVerifyRequestDelay: man.getConfigDuration("server.vpp_verify_request_delay"),
},
Auth: AuthConfig{
BcryptCost: man.getConfigInt("auth.bcrypt_cost"),
+130
View File
@@ -0,0 +1,130 @@
package config
import (
"context"
"errors"
"fmt"
"math/rand"
"strings"
"time"
aws_config "github.com/aws/aws-sdk-go-v2/config"
"github.com/aws/aws-sdk-go-v2/service/secretsmanager"
"github.com/aws/aws-sdk-go-v2/service/secretsmanager/types"
"github.com/fleetdm/fleet/v4/server/aws_common"
)
// SecretsManagerClient interface for dependency injection and testing
type SecretsManagerClient interface {
GetSecretValue(ctx context.Context, params *secretsmanager.GetSecretValueInput,
optFns ...func(*secretsmanager.Options)) (*secretsmanager.GetSecretValueOutput, error)
}
// parseRegionFromSecretARN extracts the AWS region from a Secrets Manager ARN
func parseRegionFromSecretARN(arn string) (string, error) {
// ARN format: arn:aws:secretsmanager:region:account:secret:name
parts := strings.Split(arn, ":")
if len(parts) < 6 || parts[0] != "arn" || parts[1] != "aws" || parts[2] != "secretsmanager" {
return "", fmt.Errorf("invalid Secrets Manager ARN format: %s", arn)
}
region := parts[3]
if region == "" {
return "", fmt.Errorf("region not found in ARN: %s", arn)
}
return region, nil
}
// retrieveSecretWithRetry retrieves the secret from AWS with retry logic
func retrieveSecretWithRetry(ctx context.Context, client SecretsManagerClient, secretArn string) (string, error) {
const maxRetries = 3
var lastErr error
for attempt := 0; attempt < maxRetries; attempt++ {
if attempt > 0 {
// Exponential backoff with jitter: base 100ms with ±50% randomization
baseBackoff := time.Duration(100*(1<<uint(attempt-1))) * time.Millisecond // #nosec G115 - attempt is bounded by maxRetries
jitter := time.Duration(rand.Float64()*float64(baseBackoff)) - baseBackoff/2 // #nosec G404 - not security sensitive
backoff := baseBackoff + jitter
select {
case <-ctx.Done():
return "", ctx.Err()
case <-time.After(backoff):
}
}
input := &secretsmanager.GetSecretValueInput{
SecretId: &secretArn,
}
output, err := client.GetSecretValue(ctx, input)
if err != nil {
lastErr = err
// Don't retry certain errors
var notFoundErr *types.ResourceNotFoundException
var unauthorizedErr *types.InvalidRequestException
var invalidParamErr *types.InvalidParameterException
if errors.As(err, &notFoundErr) {
return "", fmt.Errorf("secret not found: %s", secretArn)
}
if errors.As(err, &unauthorizedErr) {
return "", fmt.Errorf("access denied to secret: %s", secretArn)
}
if errors.As(err, &invalidParamErr) {
return "", fmt.Errorf("invalid secret ARN: %s", secretArn)
}
// Retry for other errors (network issues, throttling, etc.)
continue
}
// Extract secret value
if output.SecretString != nil {
return *output.SecretString, nil
}
if output.SecretBinary != nil {
return "", fmt.Errorf("secret %s contains binary data, expected string", secretArn)
}
return "", fmt.Errorf("secret %s contains no data", secretArn)
}
return "", fmt.Errorf("failed to retrieve secret after %d attempts: %w", maxRetries, lastErr)
}
// RetrieveSecretsManagerSecret retrieves a secret from AWS Secrets Manager
// with support for STS assume role authentication
func RetrieveSecretsManagerSecret(ctx context.Context, secretArn, assumeRoleArn, externalID string) (string, error) {
return RetrieveSecretsManagerSecretWithOptions(ctx, secretArn, assumeRoleArn, externalID)
}
// RetrieveSecretsManagerSecretWithOptions retrieves a secret from AWS Secrets Manager
// with custom AWS config options (useful for testing with LocalStack)
func RetrieveSecretsManagerSecretWithOptions(ctx context.Context, secretArn, assumeRoleArn, externalID string, opts ...func(*aws_config.LoadOptions) error) (string, error) {
region, err := parseRegionFromSecretARN(secretArn)
if err != nil {
return "", fmt.Errorf("invalid secret ARN: %w", err)
}
configOpts := []func(*aws_config.LoadOptions) error{aws_config.WithRegion(region)}
configOpts = append(configOpts, opts...)
cfg, err := aws_config.LoadDefaultConfig(ctx, configOpts...)
if err != nil {
return "", fmt.Errorf("failed to load AWS config: %w", err)
}
if assumeRoleArn != "" {
cfg, err = aws_common.ConfigureAssumeRoleProvider(cfg, nil, assumeRoleArn, externalID)
if err != nil {
return "", fmt.Errorf("failed to configure assume role: %w", err)
}
}
client := secretsmanager.NewFromConfig(cfg)
return retrieveSecretWithRetry(ctx, client, secretArn)
}
+271
View File
@@ -0,0 +1,271 @@
package config
import (
"context"
"errors"
"fmt"
"os"
"testing"
"github.com/aws/aws-sdk-go-v2/aws"
aws_config "github.com/aws/aws-sdk-go-v2/config"
"github.com/aws/aws-sdk-go-v2/credentials"
"github.com/aws/aws-sdk-go-v2/service/secretsmanager"
"github.com/aws/aws-sdk-go-v2/service/secretsmanager/types"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/mock"
"github.com/stretchr/testify/require"
)
// mockSecretsManagerClient is a mock implementation of the SecretsManagerClient interface
type mockSecretsManagerClient struct {
mock.Mock
}
// GetSecretValue mocks the AWS Secrets Manager GetSecretValue operation
func (m *mockSecretsManagerClient) GetSecretValue(ctx context.Context, params *secretsmanager.GetSecretValueInput, optFns ...func(*secretsmanager.Options)) (*secretsmanager.GetSecretValueOutput, error) {
args := m.Called(ctx, params)
if args.Get(0) == nil {
return nil, args.Error(1)
}
return args.Get(0).(*secretsmanager.GetSecretValueOutput), args.Error(1)
}
func TestParseRegionFromARN(t *testing.T) {
testCases := []struct {
name string
arn string
expectedReg string
expectError bool
}{
{
name: "valid ARN",
arn: "arn:aws:secretsmanager:us-west-2:123456789012:secret:fleet-private-key-abc123",
expectedReg: "us-west-2",
expectError: false,
},
{
name: "valid ARN different region",
arn: "arn:aws:secretsmanager:eu-central-1:123456789012:secret:my-secret-def456",
expectedReg: "eu-central-1",
expectError: false,
},
{
name: "invalid ARN format",
arn: "invalid-arn-format",
expectedReg: "",
expectError: true,
},
{
name: "wrong service",
arn: "arn:aws:s3:us-west-2:123456789012:bucket/my-bucket",
expectedReg: "",
expectError: true,
},
{
name: "empty region",
arn: "arn:aws:secretsmanager::123456789012:secret:my-secret",
expectedReg: "",
expectError: true,
},
}
for _, tc := range testCases {
t.Run(tc.name, func(t *testing.T) {
region, err := parseRegionFromSecretARN(tc.arn)
if tc.expectError {
assert.Error(t, err)
assert.Empty(t, region)
} else {
assert.NoError(t, err)
assert.Equal(t, tc.expectedReg, region)
}
})
}
}
func TestRetrieveSecretWithRetry_Success(t *testing.T) {
mockClient := &mockSecretsManagerClient{}
expectedKey := "test-32-byte-key-for-aes-encryption"
secretArn := "arn:aws:secretsmanager:us-west-2:123456789012:secret:test-secret" // #nosec G101 - test data
mockClient.On("GetSecretValue", mock.Anything, mock.MatchedBy(func(input *secretsmanager.GetSecretValueInput) bool {
return input.SecretId != nil && *input.SecretId == secretArn
})).Return(&secretsmanager.GetSecretValueOutput{
SecretString: &expectedKey,
}, nil)
key, err := retrieveSecretWithRetry(context.Background(), mockClient, secretArn)
require.NoError(t, err)
assert.Equal(t, expectedKey, key)
mockClient.AssertExpectations(t)
}
func TestRetrieveSecretWithRetry_BinarySecret(t *testing.T) {
mockClient := &mockSecretsManagerClient{}
secretArn := "arn:aws:secretsmanager:us-west-2:123456789012:secret:test-secret" // #nosec G101 - test data
binaryData := []byte("binary-data")
mockClient.On("GetSecretValue", mock.Anything, mock.MatchedBy(func(input *secretsmanager.GetSecretValueInput) bool {
return input.SecretId != nil && *input.SecretId == secretArn
})).Return(&secretsmanager.GetSecretValueOutput{
SecretBinary: binaryData,
}, nil)
_, err := retrieveSecretWithRetry(context.Background(), mockClient, secretArn)
require.Error(t, err)
assert.Contains(t, err.Error(), "contains binary data, expected string")
mockClient.AssertExpectations(t)
}
func TestRetrieveSecretWithRetry_EmptySecret(t *testing.T) {
mockClient := &mockSecretsManagerClient{}
secretArn := "arn:aws:secretsmanager:us-west-2:123456789012:secret:test-secret" // #nosec G101 - test data
mockClient.On("GetSecretValue", mock.Anything, mock.MatchedBy(func(input *secretsmanager.GetSecretValueInput) bool {
return input.SecretId != nil && *input.SecretId == secretArn
})).Return(&secretsmanager.GetSecretValueOutput{
// Both SecretString and SecretBinary are nil
}, nil)
_, err := retrieveSecretWithRetry(context.Background(), mockClient, secretArn)
require.Error(t, err)
assert.Contains(t, err.Error(), "contains no data")
mockClient.AssertExpectations(t)
}
func TestRetrieveSecretWithRetry_ErrorHandling(t *testing.T) {
testCases := []struct {
name string
mockError error
expectedErr string
shouldRetry bool
}{
{
name: "ResourceNotFound",
mockError: &types.ResourceNotFoundException{},
expectedErr: "secret not found",
shouldRetry: false,
},
{
name: "InvalidRequest",
mockError: &types.InvalidRequestException{},
expectedErr: "access denied",
shouldRetry: false,
},
{
name: "InvalidParameter",
mockError: &types.InvalidParameterException{},
expectedErr: "invalid secret ARN",
shouldRetry: false,
},
{
name: "NetworkError",
mockError: errors.New("network timeout"),
expectedErr: "failed to retrieve secret after 3 attempts",
shouldRetry: true,
},
}
for _, tc := range testCases {
t.Run(tc.name, func(t *testing.T) {
mockClient := &mockSecretsManagerClient{}
secretArn := "arn:aws:secretsmanager:us-west-2:123456789012:secret:test-secret" // #nosec G101 - test data, not real credentials
if tc.shouldRetry {
// Should be called 3 times for retryable errors
mockClient.On("GetSecretValue", mock.Anything, mock.Anything).Return(
(*secretsmanager.GetSecretValueOutput)(nil), tc.mockError).Times(3)
} else {
// Should only be called once for non-retryable errors
mockClient.On("GetSecretValue", mock.Anything, mock.Anything).Return(
(*secretsmanager.GetSecretValueOutput)(nil), tc.mockError).Once()
}
_, err := retrieveSecretWithRetry(context.Background(), mockClient, secretArn)
require.Error(t, err)
assert.Contains(t, err.Error(), tc.expectedErr)
mockClient.AssertExpectations(t)
})
}
}
func TestRetrieveSecretWithRetry_ContextCancellation(t *testing.T) {
mockClient := &mockSecretsManagerClient{}
secretArn := "arn:aws:secretsmanager:us-west-2:123456789012:secret:test-secret" // #nosec G101 - test data, not real credentials
// Create a context that's already cancelled
ctx, cancel := context.WithCancel(context.Background())
cancel()
// Mock the first call to return a retryable error
mockClient.On("GetSecretValue", mock.Anything, mock.Anything).Return(
(*secretsmanager.GetSecretValueOutput)(nil), errors.New("network error")).Once()
_, err := retrieveSecretWithRetry(ctx, mockClient, secretArn)
require.Error(t, err)
assert.Equal(t, context.Canceled, err)
mockClient.AssertExpectations(t)
}
func TestRetrieveSecretsManagerSecret_LocalStack(t *testing.T) {
if os.Getenv("LOCALSTACK_URL") == "" {
t.Skip("LOCALSTACK_URL not set, skipping LocalStack integration test")
}
ctx := context.Background()
localStackURL := os.Getenv("LOCALSTACK_URL")
if localStackURL == "" {
localStackURL = "http://localhost:4566"
}
localStackOpts := []func(*aws_config.LoadOptions) error{
aws_config.WithBaseEndpoint(localStackURL),
aws_config.WithCredentialsProvider(credentials.NewStaticCredentialsProvider("test", "test", "")),
}
// Configure LocalStack client
cfg, err := aws_config.LoadDefaultConfig(ctx, append(localStackOpts, aws_config.WithRegion("us-east-1"))...)
require.NoError(t, err)
client := secretsmanager.NewFromConfig(cfg)
secretName := "fleet-test-private-key-localstack"
privateKey := "test-key-exactly-32-bytes-long!"
secretArn := fmt.Sprintf("arn:aws:secretsmanager:us-east-1:000000000000:secret:%s", secretName)
// Clean up any existing secret
_, _ = client.DeleteSecret(ctx, &secretsmanager.DeleteSecretInput{
SecretId: &secretName,
ForceDeleteWithoutRecovery: aws.Bool(true),
})
_, err = client.CreateSecret(ctx, &secretsmanager.CreateSecretInput{
Name: &secretName,
SecretString: &privateKey,
Description: aws.String("password"),
})
require.NoError(t, err, "Failed to create secret in LocalStack")
// Clean up after test
defer func() {
_, _ = client.DeleteSecret(ctx, &secretsmanager.DeleteSecretInput{
SecretId: &secretName,
ForceDeleteWithoutRecovery: aws.Bool(true),
})
}()
retrievedKey, err := RetrieveSecretsManagerSecretWithOptions(ctx, secretArn, "", "", localStackOpts...)
require.NoError(t, err)
assert.Equal(t, privateKey, retrievedKey)
// Test with invalid ARN
invalidArn := "arn:aws:secretsmanager:us-east-1:000000000000:secret:nonexistent-secret"
_, err = RetrieveSecretsManagerSecretWithOptions(ctx, invalidArn, "", "", localStackOpts...)
require.Error(t, err)
assert.Contains(t, err.Error(), "secret not found")
}