Cache secrets with Vault Agent
Nearly all requests to Vault must be accompanied by a valid token which means that Vault clients must first authenticate with Vault and acquire a token. Secure Introduction of Vault Clients tutorial talked about the approaches to solve this secret zero problem.
Vault Agent with AWS and Vault Agent with Kubernetes tutorials walked through the Vault Agent's Auto-Auth.
Note
This tutorial focuses on the Caching feature of Vault Agent.
Challenge and solution
Depending on the location of your Vault clients and its secret access frequency, you may face some scaling or latency challenge. Even with Vault Performance Replication enabled, the pressure on the storage backend increases as the number of token or lease generation requests increase. Vault 1.0 introduced batch tokens as a solution to relieve some pressure on the storage backend. By design, batch tokens do not support the same level of flexibility and features as service tokens. Therefore, if you need a periodic token for example, you would need service tokens.
To increase the availability of tokens and secrets to the clients, Vault Agent introduced the Caching function.
Vault Agent Caching can cache the tokens and leased secrets proxied through the agent which includes the auto-auth token. This allows for easier access to Vault secrets for edge applications, reduces the I/O burden for basic secrets access for Vault clusters, and allows for secure local access to leased secrets for the life of a valid token.
Prerequisites
To complete this section of the tutorial, you will need:
- AWS account and associated credentials that allow for the creation of resources
- Terraform v0.15.2 or later installed
Provision the cloud resources
Clone or download the demo assets from the hashicorp/vault-guides GitHub repository to perform the steps described in this tutorial.
Clone the repository:
$ git clone https://github.com/hashicorp/vault-guides
Or download the repository:
This repository contains supporting content for all of the Vault learn tutorials. The content specific to this tutorial can be found within a sub-directory.
Set your working directory to where the
/identity/vault-agent-caching/terraform-aws
was downloaded.$ cd vault-guides/identity/vault-agent-caching/terraform-aws
The directory contains the following files.
$ tree.├── aws.tf├── iam.tf├── instances.tf├── kms.tf├── network.tf├── outputs.tf├── security-groups.tf├── templates│ ├── userdata-vault-client.tpl│ └── userdata-vault-server.tpl├── terraform.tfvars.example├── variables.tf└── versions.tf 1 directory, 12 files
Note
The example Terraform in this repository is created for the demo purpose.
Set an
AWS_ACCESS_KEY_ID
environment variable to hold your AWS access key ID.$ export AWS_ACCESS_KEY_ID = "<YOUR_AWS_ACCESS_KEY_ID>"
Set an
AWS_SECRET_ACCESS_KEY
environment variable to hold your AWS secret access key.$ export AWS_SECRET_ACCESS_KEY = "<YOUR_AWS_SECRET_ACCESS_KEY>"
Create a file named
terraform.tfvars
and specify thekey_name
. (You can copy theterraform.tfvars.example
file and rename it asterraform.tfvars
.)Example:
terraform.tfvars
# EC2 key pair name to access EC2 instances (should already exist) on the AWS regionkey_name = "vault-test"
Tip
If you don't have an EC2 key pair, follow the AWS documentation to create one.
Perform a
terraform init
to pull down the necessary provider resources.$ terraform init
Run
terraform apply
to provision AWS resources.$ terraform apply
When prompted, enter
yes
to proceed.The Terraform output will display the public IP address to SSH into your Vault server and client instances.
...Apply complete! Resources: 23 added, 0 changed, 0 destroyed.Outputs:endpoints =Vault Server IP (public): 54.219.129.15Vault Server IP (private): 10.0.101.50For example: ssh -i vault-test.pem ubuntu@54.219.129.15Vault Client IP (public): 54.183.212.51Vault Client IP (private): 10.0.101.209For example: ssh -i vault-test.pem ubuntu@54.183.212.51
Configure AWS IAM auth method
Note
This step should be performed on the Vault Server instance.
In this step, you will configure Vault to allow AWS IAM authentication from specific IAM roles.
SSH into the Vault Server instance.
When you are prompted, enter "yes" to continue.
$ ssh -i <path_to_key> ubuntu@<public_ip_of_server>...Are you sure you want to continue connecting (yes/no)? yes
NOTE: If you received
Permissions 0664 for '<key_name>.pem' are too open
error, be sure to set the file permission appropriately.$ chmod 600 <key_name>.pem
Run the
vault operator init
command to initialize the Vault server.$ vault operator init...Initial Root Token: s.20JnHBY66EKTj9zyR6SjTMNq Success! Vault is initialized...
Note
The Vault server is configured to auto-unseal with AWS Key Management Service (KMS); therefore, once the server is initialized, it gets automatically unsealed. To learn how to auto-unseal Vault using AWS KMS, refer to the Auto-unseal using AWS KMS tutorial.
Copy the Initial Root Token value.
Log into Vault using the generated initial root token.
$ vault loginToken (will be hidden):
Examine and then execute the
/home/ubuntu/aws_auth.sh
script.$ cat aws_auth.sh
Execute the script.
$ ./aws_auth.sh Success! Enabled the kv secrets engine at: kv/Success! Uploaded policy: myappSuccess! Enabled aws auth method at: aws/Success! Data written to: auth/aws/config/clientSuccess! Data written to: auth/aws/role/app-roleSuccess! Enabled userpass auth method at: userpass/Success! Data written to: auth/userpass/users/student
This script enables key/value v1 secrets engine at
kv
, createmyapp
policy, enablesaws
auth method and create a role namedapp-role
. Also, userpass is enabled and created a user (student
) withmyapp
policy attached.Check the
myapp
policy.$ vault policy read myapp path "kv/*" { capabilities = ["create", "read", "update", "delete"]}path "aws/creds/*" { capabilities = ["read", "update"]}path "sys/leases/*" { capabilities = ["create", "update"]}path "auth/token/*" { capabilities = ["create", "update"]}
Examine and then execute the
/home/ubuntu/aws_secrets.sh
script.$ cat aws_secrets.sh
Execute the script.
$ ./aws_secrets.shSuccess! Enabled the aws secrets engine at: aws/Success! Data written to: aws/config/rootSuccess! Data written to: aws/config/leaseSuccess! Data written to: aws/roles/readonly
This script enables
aws
secrets engine, configures it and set the generated secret's lease to 1 hour and lease max to 24 hours. It also configuresreadonly
role which is mapped to is mapped toarn:aws:iam::aws:policy/AmazonS3ReadOnlyAccess
IAM policy.
Start Vault Agent
Note
This step should be performed on the Vault Client instance.
Now, SSH into the Vault Client instance.
When you are prompted, enter "yes" to continue.
$ ssh -i <path_to_key> ubuntu@<public_ip_of_client>...Are you sure you want to continue connecting (yes/no)? yes
Explore the Vault Agent configuration file,
/home/ubuntu/vault-agent.hcl
$ cat vault-agent.hcl exit_after_auth = falsepid_file = "./pidfile" auto_auth { method "aws" { mount_path = "auth/aws" config = { type = "iam" role = "app-role" } } sink "file" { config = { path = "/home/ubuntu/vault-token-via-agent" } }} cache { } api_proxy { use_auto_auth_token = true} listener "tcp" { address = "127.0.0.1:8200" tls_disable = true} vault { address = "http://<vault-server-host>:8200"}
Tip
Also, read the Server Side Request Forgery (SSRF) Protection section.
Note
Notice the
api_proxy
block sets theuse_auto_auth_token
parameter to true. Since the auto-auth sink is set to/home/ubuntu/vault-token-via-agent
, in absence of client token on requests sent to the API proxy, Vault Agent will use the auto-auth token read fromvault-token-via-agent
. Thevault
block defines the Vault server location (e.g.http://10.0.101.209:8200
).Execute the following command to start the Vault Agent with log level set to
debug
.$ vault agent -config=/home/ubuntu/vault-agent.hcl -log-level=debug ==> Vault server started! Log data will stream in below: ==> Vault agent configuration: Api Address 1: http://127.0.0.1:8200 Cgo: disabled Log Level: debug Version: Vault v1.3.0 Version Sha: 30f07c76e1ea0551f28f5b8f7537f0de49d27f80 2019-03-06T03:12:47.616Z [INFO] sink.file: creating file sink2019-03-06T03:12:47.616Z [INFO] sink.file: file sink configured: path=/home/ubuntu/vault-token-via-agent2019-03-06T03:12:47.617Z [DEBUG] cache: auto-auth token is allowed to be used; configuring inmem sink2019-03-06T03:12:47.617Z [INFO] cache: starting listener: addr=127.0.0.1:82002019-03-06T03:12:47.618Z [INFO] auth.handler: starting auth handler2019-03-06T03:12:47.618Z [INFO] auth.handler: authenticating2019-03-06T03:12:47.618Z [INFO] sink.server: starting sink server2019-03-06T03:12:47.883Z [INFO] auth.handler: authentication successful, sending token to sinks2019-03-06T03:12:47.883Z [INFO] auth.handler: starting renewal process2019-03-06T03:12:47.883Z [INFO] sink.file: token written: path=/home/ubuntu/vault-token-via-agent2019-03-06T03:12:47.883Z [DEBUG] cache.leasecache: storing auto-auth token into the cache2019-03-06T03:12:47.887Z [INFO] auth.handler: renewed auth token
Open a second SSH terminal into the client machine.
Verify the client token stored in
/home/ubuntu/vault-token-via-agent
.$ more vault-token-via-agent
Check the details about the token (e.g. attached policies, TTL, etc.).
$ VAULT_TOKEN="$(cat vault-token-via-agent)" vault token lookup
Set
VAULT_AGENT_ADDR
environment variable.$ export VAULT_AGENT_ADDR="http://127.0.0.1:8200"
Request dynamic secrets and tokens via agent
Execute the following command to request an AWS credential.
$ vault read aws/creds/readonly Key Value--- -----lease_id aws/creds/readonly/oWBfLT2nIhXq9ZgWaC2e1fHylease_duration 1hlease_renewable trueaccess_key AKIAJMRXXXXXXXXMAGIAsecret_key swNXag6aVaultHappyYayAaivQ1iwMBaKCvuDoIsecurity_token <nil>
See the logs in the terminal where Vault Agent is running.
...[INFO] cache: received request: path=/v1/aws/creds/readonly method=GET[DEBUG] cache: using auto auth token: path=/v1/aws/creds/readonly method=GET[DEBUG] cache.leasecache: forwarding request: path=/v1/aws/creds/readonly method=GET[INFO] cache.apiproxy: forwarding request: path=/v1/aws/creds/readonly method=GET[DEBUG] cache.leasecache: processing lease response: path=/v1/aws/creds/readonly method=GET[DEBUG] cache.leasecache: storing response into the cache: path=/v1/aws/creds/readonly method=GET[DEBUG] cache.leasecache: initiating renewal: path=/v1/aws/creds/readonly method=GET[DEBUG] cache.leasecache: secret renewed: path=/v1/aws/creds/readonly
The log indicates that auto-auth token was used to connect Vault. The received request was forwarded to the Vault server, and the returned response was cached (find an entry
storing response into the cache
in the log).Benefit of Vault Agent
Vault Agent will manage the lifecycle of cached tokens and leases automatically so that the clients do not need to implement a logic to renew the tokens and leases.
Re-run the
vault
command to study the agent behavior.$ vault read aws/creds/readonly Key Value--- -----lease_id aws/creds/readonly/oWBfLT2nIhXq9ZgWaC2e1fHylease_duration 1hlease_renewable trueaccess_key AKIAJMRXXXXXXXXMAGIAsecret_key swNXag6aVaultHappyYayAaivQ1iwMBaKCvuDoIsecurity_token <nil>
The agent log indicates:
[INFO] cache: received request: path=/v1/aws/creds/readonly method=GET[DEBUG] cache: using auto auth token: path=/v1/aws/creds/readonly method=GET[DEBUG] cache.leasecache: returning cached response: path=/v1/aws/creds/readonly
Tip
This returns the same AWS credentials. Notice that the
request_id
andlease_id
match to the ones you received the first time.Log in as a user,
student
via agent whose password is "pAssw0rd
".$ vault login -method=userpass username="student" password="pAssw0rd" Success! You are now authenticated. The token information displayed belowis already stored in the token helper. You do NOT need to run "vault login"again. Future Vault requests will automatically use this token. Key Value--- -----token s.AKsRLji9mPq4xNZP1UIJggk6token_accessor lU69kHGAPtYMoQndeCSoYuhstoken_duration 48htoken_renewable truetoken_policies ["default" "myapp"]...snip...
Store the acquired token in VAULT_TOKEN environment variable.
$ export VAULT_TOKEN="s.AKsRLji9mPq4xNZP1UIJggk6"
Examine the agent log in the other terminal.
...[INFO] cache: received request: path=/v1/auth/userpass/login/student method=POST[DEBUG] cache: using auto auth token: path=/v1/auth/userpass/login/student method=POST[DEBUG] cache.leasecache: forwarding request: path=/v1/auth/userpass/login/student method=POST[INFO] cache.apiproxy: forwarding request: path=/v1/auth/userpass/login/student method=POST[DEBUG] cache.leasecache: processing auth response: path=/v1/auth/userpass/login/student method=POST[DEBUG] cache.leasecache: storing response into the cache: path=/v1/auth/userpass/login/student method=POST[DEBUG] cache.leasecache: initiating renewal: path=/v1/auth/userpass/login/student method=POST[DEBUG] cache.leasecache: secret renewed: path=/v1/auth/userpass/login/student...
The log should indicate that the acquired token for user
student
is successfully cached ([DEBUG] cache.leasecache: storing response into the cache
). This is because thelogin
request was sent through the agent.Create a new token with TTL set to 12 minutes.
$ vault token create -ttl=12m Key Value--- -----token s.e1h5wPbZ7i6FLCjPDqpsfDBGtoken_accessor y8K7xJr15fFjSowWW1t3VERTtoken_duration 12m...
Examine the agent log in the other terminal.
...[INFO] cache: received request: path=/v1/auth/token/create method=POST[DEBUG] cache.leasecache: forwarding request: path=/v1/auth/token/create method=POST[INFO] cache.apiproxy: forwarding request: path=/v1/auth/token/create method=POST[DEBUG] cache.leasecache: processing auth response: path=/v1/auth/token/create method=POST[DEBUG] cache.leasecache: setting parent context: path=/v1/auth/token/create method=POST[DEBUG] cache.leasecache: storing response into the cache: path=/v1/auth/token/create method=POST...
This time, instead of using the auto-auth token, the command is using the
student
token generated upon successful login. The resulting token was successfully cached.NOTE: The agent will automatically renew the token to keep it valid:
[DEBUG] cache.leasecache: renewal received; updating cache: path=/v1/auth/token/create
Re-send the token create request.
$ vault token create -ttl=12m Key Value--- -----token s.e1h5wPbZ7i6FLCjPDqpsfDBGtoken_accessor y8K7xJr15fFjSowWW1t3VERT...
The same token should be returned, and in the agent log, find the following:
[INFO] cache: received request: path=/v1/auth/token/create method=POST[DEBUG] cache.leasecache: returning cached response: path=/v1/auth/token/create
Cache evictions
The eviction of cache will occur when the agent fails to renew leases or tokens. This can happen when the cached lease/token hits it's maximum TTL or if the renewal results in an error. Agent also does some best-effort cache evictions by observing specific request types and response codes.
While agent observes requests and evicts cached entries automatically, you can
trigger a cache eviction by invoking the /agent/v1/cache-clear
endpoint.
If you need to manually evict a stale lease, invoke the
/agent/v1/cache-clear
endpoint with lease ID for which you wish to evict from the cache.Example:
$ curl --request POST \ --data '{ "type": "lease", "value": "aws/creds/readonly/8DLXo8nsgcIpx7iQPpCvWtlc" }' \ $VAULT_AGENT_ADDR/agent/v1/cache-clear
The agent log should show the following:
[DEBUG] cache.leasecache: received cache-clear request: type=lease namespace= value=aws/creds/readonly/8DLXo8nsgcIpx7iQPpCvWtlc[DEBUG] cache.leasecache: canceling context of index attached to accessor[DEBUG] cache.leasecache: successfully cleared matching cache entries[DEBUG] cache.leasecache: context cancelled; stopping renewer: path=/v1/aws/creds/readonly[DEBUG] cache.leasecache: evicting index from cache: id=dd0f9e775... path=/v1/aws/creds/readonly method=GET
Let's see what happens when you revoke a token.
$ vault token revoke s.AkdB7Rgr0t8mIexWEGHPsNfn
Or, revoke a token via API.
$ curl --request POST \ --data '{"token": "s.AkdB7Rgr0t8mIexWEGHPsNfn"}' \ $VAULT_AGENT_ADDR/v1/auth/token/revoke
Examine the agent log:
...[INFO] cache: received request: path=/v1/auth/token/revoke method=POST[DEBUG] cache: using auto auth token: path=/v1/auth/token/revoke method=POST[DEBUG] cache.leasecache: forwarding request: path=/v1/auth/token/revoke method=POST[INFO] cache.apiproxy: forwarding request: path=/v1/auth/token/revoke method=POST[DEBUG] cache.leasecache: cancelling context of index attached to token[DEBUG] cache.leasecache: successfully cleared matching cache entries[DEBUG] cache.leasecache: triggered caching eviction from revocation request[DEBUG] cache.leasecache: context cancelled; stopping renewer: path=/v1/auth/token/create[DEBUG] cache.leasecache: evicting index from cache: id=b5715bdca771174... path=/v1/auth/token/create method=POST
Note
In the agent log, notice the message, "triggered caching eviction from revocation request". When a token revocation request is made via the agent, the agent evicts the cached entries associated with the revoked token.
Similarly, if you revoked a staled AWS lease, the agent will automatically evict the cache.
If a situation requires you to clear all cached tokens and leases (e.g. reset after a number of testing), set the
type
toall
.$ curl --request POST --data '{ "type": "all" }' \ $VAULT_AGENT_ADDR/agent/v1/cache-clear
Clean up
On the server SSH terminal, execute the following command to revoke all leases.
$ vault lease revoke -prefix aws/creds/readonly
Execute the following commands to destroy cloud resources.
$ terraform apply -destroy -auto-approve
Delete state files.
$ rm -rf .terraform terraform.tfstate*
Server side request forgery (SSRF) protection
Note
Vault 1.3 introduced Server Side Request Forgery (SSRF) protection for Vault Agent. To leverage this feature, download Vault 1.3 or later.
Add require_request_header = true
in the Agent's listener configuration
stanza to enable the SSRF protection.
Example:
exit_after_auth = falsepid_file = "./pidfile" auto_auth { method "aws" { mount_path = "auth/aws" config = { type = "iam" role = "app-role" } } sink "file" { config = { path = "/home/ubuntu/vault-token-via-agent" } }} cache { } api_proxy { use_auto_auth_token = true} listener "tcp" { address = "127.0.0.1:8200" require_request_header = true} vault { address = "http://<vault-server-host>:8200"}
Notice the listener
block.
When the require_request_header
is set to true
, the Agent listener will
reject all requests that do NOT have the proper X-Vault-Request: true
header.
For example, the command to request AWS credentials you saw in the request dynamic secrets and tokens via agent step must look like:
$ curl -header "X-Vault-Request: true" \ $VAULT_AGENT_ADDR/v1/aws/creds/readonly
With absence of the X-Vault-Request
header, the Agent will throw missing 'X-Vault-Request' header
error, and the request will not be propagated to the
Vault server.
Help and reference
- Blog post: Why Use the Vault Agent for Secrets Management?
- Video: Streamline Secrets Management with Vault Agent and Vault 0.11
- Secure Introduction of Vault Clients
- Vault Agent Auto-Auth
- Vault Agent Caching
- AWS Auth Method