Skip to main content
Version: current [26.x]

Deploy Dremio on Kubernetes

You can follow these instructions to deploy Dremio on Kubernetes provisioned through a cloud provider or running in an on-premises environment.

FREE TRIAL

If you are using an Enterprise Edition free trial, go to Get Started with the Enterprise Edition Free Trial.

Prerequisites​

Before deploying Dremio on Kubernetes, ensure you have the following:

  • A hosted Kubernetes environment to deploy and manage the Dremio cluster.
    Each Dremio release is tested against Amazon Elastic Kubernetes Service (EKS), Azure Kubernetes Engine (GKE), and Google Kubernetes Engines (GKE) to ensure compatibility. If you have a containerization platform built on top of Kubernetes that is not listed here, please contact your provider and the Dremio Account Team regarding compatibility.
  • Helm 3 installed on your local machine to run Helm commands. For installation instructions, refer to Installing Helm in the Helm documentation.
  • A local kubectl configured to access your Kubernetes cluster. For installation instructions, refer to kubectl in the Kubernetes documentation.
  • Object Storage: Amazon S3 (including S3-compatible, e.g., MinIO), Azure Storage, or Google Cloud Storage (GCS).
  • Storage classes that support ReadWriteOnce (RWO) access mode and ideally can create expandable volumes.

Additional Prerequisites for the Enterprise Edition​

For the Enterprise Edition, you must:

  • Create an account on Quay.io to access Dremio's OCI repository, which stores Dremio's Helm charts and images.
    To get access, contact your Dremio account executive or Dremio Support.

    note

    If your internet access doesn't allow reaching Dremio's OCI repository in Quay.io, consider using a private mirror to fetch Dremio's Helm chart images.

  • Get a valid license key issued by Dremio to put in the Helm chart. To obtain the license, refer to Licensing.

Additional Prerequisites for the OpenShift​

Before deploying Dremio onto OpenShift, you additionally need the following:

  • Have the OpenShift oc CLI command configured and authenticated. For the installation instructions, see OpenShift CLI (oc).

Node Tuning for OpenSearch on OpenShift​

OpenSearch requires the vm.max_map_count kernel parameter to be set to at least 262144.

This parameter controls the maximum number of memory map areas a process can have, and OpenSearch uses memory-mapped files extensively for performance.

Without this setting, OpenSearch pods will fail to start with errors related to virtual memory limits.

Since the Helm chart sets setVMMaxMapCount: false for OpenShift compatibility (to avoid privileged init containers), you need to configure this kernel parameter at the node level. The recommended way to do it is a Node Tuning Operator. This Operator ships with OpenShift and provides a declarative way to configure kernel parameters.

Create a Tuned resource to configure the required kernel parameter:

The tuned-opensearch.yaml configuration file
apiVersion: tuned.openshift.io/v1
kind: Tuned
metadata:
  name: openshift-opensearch
  namespace: openshift-cluster-node-tuning-operator
spec:
  profile:
  - data: |
      [main]
      summary=Optimize systems running OpenSearch on OpenShift nodes
      include=openshift-node
      [sysctl]
      vm.max_map_count=262144
    name: openshift-opensearch
  recommend:
  - match:
    - label: tuned.openshift.io/opensearch
      type: pod
    priority: 20
    profile: openshift-opensearch

This YAML should be saved locally and applied to any cluster you intend to deploy Dremio:

oc apply -f tuned-opensearch.yaml

Step 1: Deploy Dremio​

To deploy the Dremio cluster in Kubernetes, do the following:

  1. Configure your values to deploy Dremio to Kubernetes in the file values-overrides.yaml. For that, go to Configuring Your Values to Deploy Dremio to Kubernetes and get back here to continue with the deployment.

  2. On your terminal, start the deployment by installing Dremio's Helm chart:

    Run the following command for any Kubernetes environment except for OpenShift:

    helm install <your-dremio-install-release> oci://quay.io/dremio/dremio-helm \
    --values <your-local-path>/values-overrides.yaml \
    --version <optional-helm-chart-version> \
    --set-file <optional-config-files> \
    --wait

    Where:

    • <your-dremio-install-release> - The name that identifies your Dremio installation. For example, dremio-1-0.
    • <your-local-path> - The path to reach your values-overrides.yaml configuration file.
    • (Optional) --version <optional-helm-chart-version> - The version of Dremio's Helm chart to be used. If not provided defaults to the latest.
    • (Optional) --set-file <optional-config-file> - An optional configuration file for deploying Dremio. For example, an Identity Provider configuration file, which is not defined in the values-overrides.yaml and can be provided here through this option.

  3. Monitor the deployment using the following commands:

    Run the following command for any Kubernetes environment except for OpenShift:

    kubectl get pods

    When all of the pods are in the Ready state, the deployment is complete.

    Troubleshooting

    • If a pod remains in Pending state for more than a few minutes, run the following command to view its status to check for issues, such as insufficient resources for scheduling:

      kubectl describe pods <pod-name>

    • If the events at the bottom of the output mention insufficient CPU or memory, do one of the following:

      • Adjust the values in the values-overrides.yaml configuration file and redeploy.

      • Add more resources to your Kubernetes cluster.

    • If a pod returns a failed state (especially dremio-master-0, the most important pod), use the following commands to collect the logs:

      Run the following command for any Kubernetes environment except for OpenShift:

      kubectl logs dremio-master-0

Step 2: Connecting to Dremio​

Now that you've installed the Helm chart and deployed Dremio on Kubernetes, the next step is connecting to Dremio, where you have the following options:

To connect to Dremio via the Dremio console, run the following command to use the services dremio-client in Kubernetes to find the host for the Dremio console:

$ kubectl get services dremio-client
NAME            TYPE           CLUSTER-IP      EXTERNAL-IP       PORT(S)                          AGE
...             ...            ...             ...               ...                              ...

  • If the value in the TYPE column of the output is LoadBalancer, access the Dremio console through the address in the EXTERNAL_IP column and port 9047.
    For example, in the output below, the value under the EXTERNAL-IP column is 8.8.8.8. Therefore, access the Dremio console through http://8.8.8.8:9047.

    $ kubectl get services dremio-client
    NAME            TYPE           CLUSTER-IP      EXTERNAL-IP       PORT(S)                          AGE
    dremio-client   LoadBalancer   10.99.227.180   8.8.8.8           31010:32260/TCP,9047:30620/TCP   2d

    If you want to change the exposed port on the load balancer, change the value of the setting coordinator.web.port in the file values-overrides.yaml.

  • If the value in the TYPE column of the output is NodePort, access the Dremio console through http://localhost:30670.