Database Access with Self-Hosted CockroachDB

Teleport can provide secure access to CockroachDB via the Teleport Database Service. This allows for fine-grained access control through Teleport's RBAC.

In this guide, you will:

1. Configure an CockroachDB with mutual TLS authentication.
2. Join the CockroachDB database to your Teleport cluster.
3. Connect to the CockroachDB database via the Teleport Database Service.

• Self-Hosted
• Teleport Enterprise Cloud

Prerequisites

• A running Teleport cluster. If you want to get started with Teleport, sign up for a free trial or set up a demo environment.

• The tctl admin tool and tsh client tool version >= 15.2.2.

  On Teleport Enterprise, you must use the Enterprise version of tctl, which you can download from your Teleport account workspace. Otherwise, visit Installation for instructions on downloading tctl and tsh for Teleport Community Edition.

• CockroachDB cluster.

• A host, e.g., an Amazon EC2 instance, where you will run the Teleport Database Service.

• To check that you can connect to your Teleport cluster, sign in with tsh login, then verify that you can run tctl commands using your current credentials. tctl is supported on macOS and Linux machines.

  For example:

  tsh login --proxy=teleport.example.com --user=[email protected]
  tctl status
  Cluster  teleport.example.com
  Version  15.2.2
  CA pin   sha256:abdc1245efgh5678abdc1245efgh5678abdc1245efgh5678abdc1245efgh5678

  If you can connect to the cluster and run the tctl status command, you can use your current credentials to run subsequent tctl commands from your workstation. If you host your own Teleport cluster, you can also run tctl commands on the computer that hosts the Teleport Auth Service for full permissions.

• A certificate authority to issue CockroachDB certificates for nodes in your CockroachDB cluster.

  Why do I need my own CA?

  Distributed databases like CockroachDB use mTLS for node-to-node communication. Teleport requires that you have your own CA to issue certificates for node-to-node mTLS communication.

  Teleport uses a split-CA architecture for database access. The Teleport db CA issues server certificates and the db_client CA issues client certificates.

  Databases are configured to trust the Teleport db_client CA for client authentication, but not the db CA. Additionally, Teleport only issues ephemeral db_client CA certificates.

  When a CockroachDB node connects to another CockroachDB node, it must present a certificate that the other node trusts for client authentication. Since Teleport does not issue long-lived db_client certificates, the node needs to have a long-lived certificate issued by another CA that its peer node trusts.

  The split db and db_client CA architecture was introduced as a security fix in Teleport versions: 13.4.17, 14.3.7, and 15.

  See Database CA Migrations for more information.

Step 1/4. Set up the Teleport Database Service

The Database Service requires a valid auth token to connect to the cluster. Generate one by running the following command against your Teleport Auth Service and save it in /tmp/token on the node that will run the Database Service:

tctl tokens add --type=db

Install and configure Teleport where you will run the Teleport Database Service:

sudo teleport db configure create \   -o file \   --token=/tmp/token \   --proxy=teleport.example.com:443 \   --name=roach \   --protocol=cockroachdb \   --uri=roach.example.com:26257 \   --labels=env=dev 

sudo teleport db configure create \   -o file \   --token=/tmp/token \   --proxy=mytenant.teleport.sh:443 \   --name=roach \   --protocol=cockroachdb \   --uri=roach.example.com:26257 \   --labels=env=dev

sudo systemctl enable teleport
sudo systemctl start teleport

sudo teleport install systemd -o /etc/systemd/system/teleport.service
sudo systemctl enable teleport
sudo systemctl start teleport

helm repo add teleport https://charts.releases.teleport.dev

helm repo update

JOIN_TOKEN=$(cat /tmp/token)
helm install teleport-kube-agent teleport/teleport-kube-agent \  --create-namespace \  --namespace teleport-agent \  --set roles=db \  --set proxyAddr=teleport.example.com:443 \  --set authToken=${JOIN_TOKEN?} \  --set "databases[0].name=roach" \  --set "databases[0].uri=roach.example.com:26257" \  --set "databases[0].protocol=cockroachdb" \  --set "labels.env=dev" \  --version 15.2.2

JOIN_TOKEN=$(cat /tmp/token)
helm install teleport-kube-agent teleport/teleport-kube-agent \  --create-namespace \  --namespace teleport-agent \  --set roles=db \  --set proxyAddr=mytenant.teleport.sh:443 \  --set authToken=${JOIN_TOKEN?} \  --set "databases[0].name=roach" \  --set "databases[0].uri=roach.example.com:26257" \  --set "databases[0].protocol=cockroachdb" \  --set "labels.env=dev" \  --version 15.2.2

kubectl -n teleport-agent get pods
NAME                    READY   STATUS    RESTARTS   AGEteleport-kube-agent-0   1/1     Running   0          32s

Step 2/4. Create a Teleport user

tctl users add \  --roles=access \  --db-users=\* \  --db-names=\* \  alice

tctl users add \  --roles=access,requester \  --db-users=\* \  --db-names=\* \  alice

For more detailed information about database access controls and how to restrict access see RBAC documentation.

Step 3/4. Configure CockroachDB

Create a CockroachDB user

Teleport uses mutual TLS authentication with CockroachDB. Client certificate authentication is available to all CockroachDB users. If you don't have one, connect to your Cockroach cluster and create it:

CREATE USER alice WITH PASSWORD NULL;

The WITH PASSWORD NULL clause prevents the user from using password auth and mandates client certificate auth.

Make sure to assign the user proper permissions within the database cluster. Refer to Create User in the CockroachDB documentation for more information.

Set up mutual TLS

To set up mutual TLS authentication, you need to make sure that:

• Teleport trusts certificates presented by CockroachDB nodes.
• CockroachDB nodes trust client certificates signed by both your CockroachDB CA and your Teleport cluster's db_client CA.

CockroachDB nodes need to trust the Teleport db_client CA so that Teleport users can authenticate to your CockroachDB cluster as clients.

The CockroachDB CA needs to be trusted by each CockroachDB node so that nodes can authenticate themselves as clients to other nodes in the CockroachDB cluster. This is because CockroachDB uses mTLS for node-to-node communication.

CERTS_DIR=/path/to/cockroachdb/certs/dir
cp "${CERTS_DIR}/ca.crt" "${CERTS_DIR}/ca-client.crt"

tctl auth export --type=db_client >> /path/to/cockroachdb/certs/dir/ca-client.crt

  databases:
  - name: "example-cockroachdb"
    protocol: "cockroachdb"
    uri: "cockroachdb.example.com:26257"
    static_labels:
      "env": "example"
    tls:
      ca_cert_file: "/path/to/your/ca.crt"

allow:
  impersonate:
    users: ["Db"]
    roles: ["Db"]

tctl auth sign \    --format=cockroachdb \    --host=roach.example.com \    --out=/path/to/cockroachdb/certs/dir \    --ttl=2190h

cockroach cert create-client node \  --certs-dir=/path/to/cockroachdb/certs/dir \  --ca-key=ca-secrets/ca-client.key

Restart your CockroachDB nodes, passing them the directory with generated secrets via the --certs-dir flag:

cockroach start \    --certs-dir=/path/to/cockroachdb/certs/dir \    # other flags...

Alternatively, if the nodes were already started with --certs-dir=/path/to/cockroachdb/certs/dir, you can send a SIGHUP signal to the cockroach process to reload certificates without restarting the node. You must send SIGHUP as the same user that started the cockroach process:

pkill -SIGHUP -x cockroach

Step 4/4. Connect

Log in to your Teleport cluster. Your CockroachDB cluster should appear in the list of available databases:

tsh login --proxy=teleport.example.com --user=alice
tsh db ls
Name  Description         Labels
----- ------------------- -------
roach Example CockroachDB env=dev

tsh login --proxy=mytenant.teleport.sh --user=alice
tsh db ls
Name  Description         Labels
----- ------------------- -------
roach Example CockroachDB env=dev

To retrieve credentials for a database and connect to it:

tsh db connect roach

You can optionally specify the database name and the user to use by default when connecting to the database server:

tsh db connect --db-user=alice roach

To log out of the database and remove credentials:

tsh db logout roach

Troubleshooting

Unimplemented client encoding error

You may encounter the unimplemented client encoding: "sqlascii" error when connecting to your CockroachDB database if your psql uses SQL_ASCII encoding.

CockroachDB supports only UTF8 client encoding. To enforce the encoding, set the following environment variable in the shell running tsh db connect:

export PGCLIENTENCODING='utf-8'

If you are connecting the CockroachDB database with Teleport Connect, add the environment variable to your shell startup scripts and restart the Teleport Connect app.

Unable to cancel a query

If you use a PostgreSQL cli client like psql, and you try to cancel a query with ctrl+c, but it doesn't cancel the query, then you need to connect using a tsh local proxy instead. When psql cancels a query, it establishes a new connection without TLS certificates, however Teleport requires TLS certificates not only for authentication, but also to route database connections.

If you enable TLS Routing in Teleport then tsh db connect will automatically start a local proxy for every connection. Alternatively, you can connect via Teleport Connect which also uses a local proxy. Otherwise, you need to start a tsh local proxy manually using tsh proxy db and connect via the local proxy.

If you have already started a long-running query in a psql session that you cannot cancel with ctrl+c, you can start a new client session to cancel that query manually:

First, find the query's process identifier (PID):

SELECT pid,usename,backend_start,query FROM pg_stat_activity WHERE state = 'active';

Next, gracefully cancel the query using its PID. This will send a SIGINT signal to the postgres backend process for that query:

SELECT pg_cancel_backend(<PID>);

You should always try to gracefully terminate a query first, but if graceful cancellation is taking too long, then you can forcefully terminate the query instead. This will send a SIGTERM signal to the postgres backend process for that query:

SELECT pg_terminate_backend(<PID>);

See the PostgreSQL documentation on admin functions for more information about the pg_cancel_backend and pg_terminate_backend functions.

Next steps

• Learn how to restrict access to certain users and databases.

• View the High Availability (HA) guide.

• Take a look at the YAML configuration reference.

• See the full CLI reference.

• CockroachDB client authentication
• CockroachDB using split CA certificates