Deploy Label Studio on Kubernetes
Deploy Label Studio on a Kubernetes Cluster using Helm 3. You can use this Helm chart to set up Label Studio for deployment onto a Kubernetes cluster and install, upgrade, and manage the application.
Your Kubernetes cluster can be self-hosted or installed somewhere such as Amazon EKS. See the Amazon tutorial on how to Deploy a Kubernetes Application with Amazon Elastic Container Service for Kubernetes for more about deploying an app on Amazon EKS.
To install Label Studio Enterprise Edition, see Deploy Label Studio Enterprise on Kubernetes. This page is specific to the community version of Label Studio.
If you want to install Label Studio on Kubernetes and you have unrestricted access to the internet from your K8s cluster, follow these steps.
- Verify that you meet the Required software prerequisites and review the capacity planning guidance.
- Prepare the Kubernetes cluster.
- Add the Helm chart repository.
- (Optional) Set up persistent storage.
- (Optional) Configure ingress.
- (Optional) Configure values.yaml.
- (Optional) Set up TLS for PostgreSQL
- Use Helm to install Label Studio on your Kubernetes cluster.
If you use a proxy to access the internet from your Kubernetes cluster, or it is airgapped from the internet, see how to Install Label Studio without public internet access.
- Kubernetes and kubectl version 1.17 or higher
- Helm version 3.6.3 or higher
This chart has been tested and confirmed to work with the NGINX Ingress Controller and cert-manager. See Set up an ingress controller for Label Studio Kubernetes deployments for more on ingress settings with Label Studio.
Your Kubernetes cluster can be self-hosted or installed somewhere such as Amazon EKS.
To plan the capacity of your Kubernetes cluster, refer to these guidelines.
Label Studio has the following default configurations for resource requests, resource limits, and replica counts:
Before you make changes to these values, familiarize yourself with the Resource Management for Pods and Containers guidelines in the Kubernetes documentation.
If you choose to make changes to these default settings, consider the following:
|For this case
|More than 10 concurrent annotators
|Adjust the requests and limits for
resources in the
|Increase fault tolerance
|Increase the number of replicas of
|Production deployment (replicas)
|Replicas equivalent or greater than the number of availability zones in your Kubernetes cluster
Before installing Label Studio, prepare the Kubernetes cluster with kubectl.
Add the Helm chart repository to easily install and update Label Studio.
- From the command line:
helm repo add heartex https://charts.heartex.com/ helm repo update heartex
- If you want, check for available versions:
helm search repo heartex/label-studio
To configure Label Studio to use TLS for end-client connections with PostgreSQL, do the following:
- Enable TLS for your PostgreSQL instance and save Root TLS certificate, client certificate and its key for the next steps.
- Create a Kubernetes secret with your certificates, replacing
<PATH_TO_CLIENT_KEY>with paths to your certificates:
kubectl create secret generic <YOUR_SECRET_NAME> --from-file=ca.crt=<PATH_TO_CA> --from-file=client.crt=<PATH_TO_CLIENT_CRT> --from-file=client.key=<PATH_TO_CLIENT_KEY>
- Update your
ls-values.yamlfile with your newly-created Kubernetes secret:
POSTGRE_SSL_MODE: verify-ca, the server is verified by checking the certificate chain up to the root certificate stored on the client. If
POSTGRE_SSL_MODE: verify-full, the server host name will be verified to make sure it matches the name stored in the server certificate. The SSL connection will fail if the server certificate cannot be verified.
verify-full is recommended in most security-sensitive environments.
- Install or upgrade Label Studio using Helm.
Use Helm to install Label Studio on your Kubernetes cluster. Provide your custom resource definitions YAML file. Specify any environment variables that you need to set for your Label Studio installation using the
--set argument with the
helm install command.
If you are deploying to a production environment, you should set the
SSRF_PROTECTION_ENABLED: true environment variable. See Secure Label Studio.
From the command line, run the following:
helm install <RELEASE_NAME> heartex/label-studio -f ls-values.yaml
After installing, check the status of the Kubernetes pod creation:
kubectl get pods
Restart your Helm release by doing the following from the command line:
- Identify the <RELEASE_NAME> of the latest Label Studio release:
2Restart the Label Studio app:
kubectl rollout restart deployment/<RELEASE_NAME>-ls-app
To upgrade Label Studio using Helm, do the following.
- Determine the latest tag version of Label Studio and add/replace the following in your
global: image: tag: "20210914.154442-d2d1935"
- After updating the values file, retrieve the latest updates for the Helm chart:
helm repo update heartex
- Run the following from the command line to upgrade your deployment:
If you want, you can specify a version from the command line:
helm upgrade <RELEASE_NAME> heartex/label-studio -f ls-values.yaml
This command overrides the tag value stored in
helm upgrade <RELEASE_NAME> heartex/label-studio -f ls-values.yaml --set global.image.tag=20210914.154442-d2d1935
ls-values.yaml. You must update the tag value when you upgrade or redeploy your instance to avoid version downgrades.
To uninstall Label Studio using Helm, delete the configuration.
From the command line, run the following:
helm delete <RELEASE_NAME>