Advanced Orchestrated container deployment (Helm)

Deploy CARTO Self-hosted using Kubernetes and Helm

Estimated time: Completing this deployment guide is expected to take approximately 3 hours. This estimate may vary based on individual familiarity with the technology stack involved and the complexity of your organization's environment.

Requirements

To deploy CARTO Self-Hosted on Kubernetes, you need:

• A CARTO Self-Hosted installation package containing your environment configuration and a license key. The package has two YAML files: carto-secrets.yaml and carto-values.yaml If you don't have it yet, you can ask for it at [e-mail].

• A domain you own, to which you can add a DNS record.

• A Kubernetes cluster. To create a cluster, see documentation on Google Cloud Platform, AWS, and Azure. This cluster must fit our hardware and software requirements for Kubernetes.

• A working installation of kubectl. To install kubectl, see documentation on Google Cloud Platform, AWS, and Azure.

• Kubectl installed on your CLI.

• A working installation of Helm v3 on version 3.6.0 or later.

• Troubleshoot.sh installed to run the preflight checks (view step 4.)

1. Add Helm repository

Type this command to add the helm repository

# Add the carto repo.
helm repo add carto https://helm.carto.com

Search for the repo to confirm you can access the carto chart.

helm search repo carto -l

2. Add your minimal customization

Domain

Create a file customizations.yaml with the domain name you want to use

appConfigValues:
  selfHostedDomain: "my.domain.com"

A full domain is required. You cannot install CARTO in a domain path like https://my.domain.com/carto

External database

At this point, we are setting up the configuration of the external database. You need to provide a PostgreSQL user with full permission on a logical database.

If you're installing in EKS and you'd like to use EKS Pod Identity to authenticate to your RDS PostgreSQL database, follow this guide.

Create the secret with both passwords:

kubectl create secret generic \
  carto-postgresql \
  --from-literal=user-password='<userPassword>' \

Edit customizations.yaml:

• user: The carto user to be created. It will be created with the previous admin user.

• user-password: The new password to be created.

• database: The database to be created.

internalPostgresql:
  # Disable the internal Postgres
  enabled: false
externalPostgresql:
  host: <YourServerIPorDNS>
  user: "carto_worskpace_admin"
  existingSecret: "carto-postgresql"
  existingSecretPasswordKey: "user-password"
  database: "carto_workspace"
  port: "5432"

SSL

In some scenarios, an SSL connection to the external database is required. In that case, you should add to customizations.yaml:

externalPostgresql:
  sslEnabled: true
  # Only applies if your PostgreSQL SSL certificate it's self-signed
    sslCA: |
    -----BEGIN CERTIFICATE-----
    ....
    -----END CERTIFICATE-----

Mutual TLS connections between the external database and the APIs are not supported, so client certificates can't be configured on your external database

3. Configure your storage buckets

CARTO Self-hosted platform needs access to some storage buckets to save some resources needed by the platform. These buckets are in charge of storing assets such as imported datasets, map snapshots and custom markers.

You can create and use your own storage buckets in any of the following supported storage providers:

• Google Cloud Storage
• AWS S3
• Azure Blob Storage

And in order to configure them, there is a detailed guide available that you should follow to complete the Self-Hosted configuration process.

4. Execute the preflight checks

Before installing CARTO Self-Hosted, you can use preflight checks to validate your configuration. Keep iterating until all checks pass—this ensures that your environment is fully prepared for installation.

As mentioned in the Requisites, you need to install the Trobleshoot.sh tools in order to run the Preflight checks. To install this tool using bash, you have to run the following commands:

1. Install the preflight and support-bundle plugins:

curl https://krew.sh/support-bundle | bash
curl https://krew.sh/preflight | bash

1. Check version

kubectl preflight version
kubectl support-bundle version 

Use the following command to run the preflight checks:

helm template \
  carto \
  carto/carto \
  -f carto-values.yaml \
  -f carto-secrets.yaml \
  -f customizations.yaml -n <NAMESPACE> | kubectl preflight -

The previous command performs validation checks to ensure that your customizations.yaml file is correctly configured for installing the CARTO platform and that you're infrastructure is ready to host a CARTO Self-Hosted installation.

Once the checks are complete, you will see the following output in your terminal:

If you encounter any issues during the preflight checks, verify the following:

• PostgreSQL Configuration: Ensure your PostgreSQL credentials are correctly set up and that the database is accessible from your cluster.

• External Proxy: If using an external proxy, confirm that it is correctly configured for connecting to external domains.
• Egress Requirements: Check that your environment meets the necessary egress requirements for CARTO Self-Hosted.
• Storage Configuration: Verify that your storage buckets are correctly set up and that the provided credentials are valid.

By addressing these areas, you can resolve common setup issues and proceed with a smooth installation.

5. Install CARTO

Use the following command to install CARTO in your Kubernetes cluster.

helm install \
  carto \
  carto/carto \
  -f carto-values.yaml \
  -f carto-secrets.yaml \
  -f customizations.yaml -n <NAMESPACE>

After installing CARTO, verify you have the required pods running by running the following command (It can take up to 5 minutes to have everything running):

kubectl get pods
NAME                                          READY   STATUS    RESTARTS   AGE
carto-accounts-www-5778cc7c7f-l42rx           1/1     Running   0          3m6s
carto-cdn-invalidator-sub-bd69b7b96-c6mmt     1/1     Running   0          3m5s
carto-http-cache-864d54db46-56xs2             1/1     Running   0          3m6s
carto-import-api-c845759bd-9m7bd              1/1     Running   0          3m7s
carto-import-worker-5bf456f684-b9mg9          1/1     Running   0          3m7s
carto-lds-api-676dcdd9c8-lqmcf                1/1     Running   0          3m6s
carto-maps-api-5699c99fb6-9dj5s               1/1     Running   0          3m4s
carto-maps-api-5699c99fb6-msfr2               1/1     Running   0          3m4s
carto-notifier-cf8f7576d-5wnm5                1/1     Running   0          3m5s
carto-redis-master-0                          1/1     Running   0          3m6s
carto-router-7578cdd848-mbrxw                 2/2     Running   0          3m5s
carto-sql-worker-64d7598c56-zt5xr             1/1     Running   0          3m6s
carto-workspace-api-66dbbdb69-cqd4p           1/1     Running   0          3m5s
carto-workspace-api-66dbbdb69-z4tj9           1/1     Running   0          3m4s
carto-workspace-subscriber-54f9cbc589-qk94m   1/1     Running   0          3m5s
carto-workspace-www-7bb5f78577-k66rr          1/1     Running   0          3m6s

Once all the services are running, verify the installation by port forwarding to localhost. Execute the following:

export CARTO_HOST_IP=127.0.0.1
export CARTO_HTTP_PORT=80
export CARTO_HTTPS_PORT=443
kubectl port-forward --namespace default svc/carto-router 80:80 443:443

Add to your /etc/hosts the domain you will use:

echo "127.0.0.1 my.domain.com" >> /etc/hosts

6. Configure an HTTPS endpoint

The entry point to the CARTO Self-Hosted is through the router service. By default, it is configured in ClusterIP mode, and it's only accessible in your machine with kubectl port-forward).

The first step here is to make the router accessible using a LoadBalancer. This provides an externally-accessible IP address that we will later configure our DNS.

Add to customizations.yaml based on the cloud you are using:

Google GKE

router:
  service:
    type: LoadBalancer

AWS EKS

router:
  service:
    type: LoadBalancer
    annotations:
      # https://kubernetes-sigs.github.io/aws-load-balancer-controller/v2.4/guide/service/nlb/
      # https://kubernetes.io/docs/concepts/services-networking/service/#ssl-support-on-aws
      service.beta.kubernetes.io/aws-load-balancer-connection-idle-timeout: "605"

Azure AKS

router:
  service:
    type: LoadBalancer
    annotations:
      # https://docs.microsoft.com/en-us/azure/aks/load-balancer-standard#additional-customizations-via-kubernetes-annotations
      service.beta.kubernetes.io/azure-load-balancer-tcp-idle-timeout: "10"

After that, upgrade the cluster:

helm upgrade carto \
  carto/carto \
  -f carto-values.yaml \
  -f carto-secrets.yaml \
  -f customizations.yaml

Get the EXTERNAL-IP of your CARTO Self-Hosted deployment with this command:

$ kubectl get svc --namespace default -w carto-router

NAME           TYPE           CLUSTER-IP   EXTERNAL-IP     PORT(S)                      AGE
carto-router   LoadBalancer   10.64.3.11   34.136.204.51   80:32339/TCP,443:32602/TCP   8h

At this point, you need to configure a DNS record that points my.domain.com to the EXTERNAL-IP (34.136.204.51).

Use your own TLS certificate

By default, the package generates a self-signed certificate with a validity of 365 days.

If you want to add your own certificate you need to create a secret:

kubectl create secret tls -n <namespace> <certificate name> \
  --cert=path/to/cert/file \
  --key=path/to/key/file

Edit customizations.yaml:

tlsCerts:
  httpsEnabled: true
  autoGenerate: false
  existingSecret:
    name: "<certificate name>"
    keyKey: "tls.key"
    certKey: "tls.crt"

Apply the changes:

helm upgrade carto \
  carto/carto \
  -f carto-values.yaml \
  -f carto-secrets.yaml \
  -f customizations.yaml

If your TLS certificate key is protected with a passphrase the CARTO Self-hosted installation won't be able to work as expected. You can easily generate a new key file without passphrase protection using the following command:

openssl rsa -in keyfile_with_passphrase.key -out new_keyfile.key

7. Post-installation checks

In order to verify CARTO Self Hosted was correctly installed, and it's functional, we recommend performing the following checks:

1. Sign in to your Self Hosted, create a user and a new organization.

2. Go to the Connections page, in the left-hand menu, create a new connection to one of the available providers.

3. Go to the Data Explorer page, click on the Upload button right next to the Connections panel. Import a dataset from a local file.

4. Go back to the Maps page, and create a new map.

5. In this new map, add a new layer from a table using the connection created in step 3.

6. Create a new layer from a SQL Query to the same table. You can use a simple query like:

SELECT * FROM <dataset_name.table_name> LIMIT 100;

1. Create a new layer from the dataset imported in step 4.

2. Make the map public, copy the sharing URL and open it in a new incognito window.

3. Go back to the Maps page, and verify your map appears there, and the map thumbnail represents the latest changes you made to the map.

Congrats! Once you've configured your custom buckets, you should have a production-ready deployment of CARTO Self-Hosted at https://my.domain.com

You may notice that the onboarding experience (demo maps, demo workflows...) and the Data Observatory-automated features (subscriptions, enrichment...) are disabled by default in your new organization, because the CARTO Data Warehouse is not enabled.

If you'd like to enable the onboarding experience and the Data Observatory features, follow the guide to enable the CARTO Data Warehouse or contact [e-mail].

If you prefer not to enable the CARTO Data Warehouse, you can still use the Data Observatory without the UI features: after getting in touch, our team can deliver the data (both premium and public subscriptions) manually to your data warehouse.

Analytics Toolbox in CARTO Self-Hosted

To fully leverage CARTO's capabilities you need to gain access to the Analytics Toolbox functions. This step is crucial to fully leverage CARTO's capabilities. Please refer to the documentation of your data warehouse provider for detailed instructions:

• Google BigQuery
• Amazon Redshift
• Snowflake
• PostgreSQL

Troubleshooting

The following standard commands of kubectl could be used to debug possible issues that might arise:

kubectl logs and kubectl describe

Database

The container workspace-migrations (included at the pod carto-workspace-api-*) will be responsible for creating a new user carto_worskpace_admin and a database carto_workspace.

To debug possible errors with the connection of the external database you might need to check the logs of this container:

kubectl logs carto-workspace-api-585f4bbd7c-qt65g -c workspace-migrations

For further assistance check our Support page.