Notebook

The Notebooks menu is used to create notebooks for data analysis, software development, or machine learning experiments using JupyterLab, which runs in Pods on the Cloudeka Deka MLOps Service Portal cluster.

Create Notebooks

Create Notebooks is used to create new notebooks on the Cloudeka Deka MLOps Service Portal. On the Notebooks page, click the + New Notebook button.

Choose Notebook: JupyterLab

The New Notebook page appears, and several fields must be completed. In the Name field, enter the name of the notebook. The name must consist of lowercase alphanumeric characters or hyphens (-), start with an alphabetic character, and end with an alphanumeric character. In this guide, JupyterLab is used as the notebook environment. In the Custom Notebook section, specify the image name created in the Deka Registry through the Service Portal Cloudeka.

Add CPU/RAM and GPUs

In the CPU/RAM section, adjust the Minimum CPU and Minimum Memory Gi values according to your requirements. In the GPUs section, specify the GPU to be used. The GPU Vendor field is automatically set to NVIDIA.

Add Workspace Volume

Workspace Volumes are used to store data used by notebooks and other workloads, helping to organize and manage data during development. In the Workspace Volume section, you can configure a volume that will be mounted to the notebook.

To view the volume configuration details, click the expand icon next to the Delete icon. The available volume types are Empty volume and Custom (Advanced). The first volume type is Empty volume, which is used to create a new storage volume that can be mounted to the notebook.

By default, the Empty volume type displays the following settings:

a. The volume type is set to Empty volume.

b. The default name for the new volume is "-volume". You can customize the name as needed.

c. The default size is 5 Gibibytes (GiB).

d. By default, the Storage Class uses storage-nvme-c1. If Use Default Class is not selected, several options are available, including Empty Storage Class, storage-mgmt, storage-nlsas, storage-nvme-c2, and other built-in classes.

e. There are three access modes available:

• ReadWriteOnce (RWO) provides exclusive read and write access to a single notebook server.

• ReadOnlyMany (ROX) provides read-only access for multiple notebook servers.

• ReadWriteMany (RWX) provides concurrent read and write access for multiple notebook servers.

f. The Mount Path is the location in the container where the volume is accessed. For Workspace Volumes, the default directory used by Jupyter Notebook is /home/jovyan.

The second volume type is Custom (Advanced). This option allows you to define a custom volume configuration using a Kubernetes volume specification. Refer to the K8s docs link for information about supported volume types and configuration options.

By default, the Custom (Advanced) type displays the following settings:

a. The volume type is set to Custom (Advanced).

b. A YAML editor is provided to define the volume configuration.

c. The Mount Path specifies the location where the volume is mounted inside the container. By default, the path is /home/jovyan.

Add Data Volume

The Data Volumes section is used to store and manage data persistently beyond the lifecycle of the notebook server, ensuring that the data remains available even if the server is stopped or deleted. In the Data Volumes section, there are two features available: Add new volume and Attach existing volume. You can add more than one data volume.

The first feature is the + Add New Volume button, which is used to create a new storage volume that can be mounted to a notebook or used by various components in a pipeline.

To view the details of the new data volume, click the expand icon next to the Delete icon. By default, the new data volume displays the following settings:

a. The volume type is set to Empty volume.

b. The default name for the new volume is "-datavol-1". You can customize the name as needed.

c. The default size is 5 Gibibytes (GiB).

d. By default, the Storage Class uses storage-nvme-c1. If Use Default Class is not selected, several options are available, including Empty Storage Class, storage-mgmt, storage-nlsas, storage-nvme-c2, and other built-in classes.

e. There are three access modes available:

• ReadWriteOnce (RWO) provides exclusive read and write access to a single notebook server.

• ReadOnlyMany (ROX) provides read-only access for multiple notebook servers.

• ReadWriteMany (RWX) provides concurrent read and write access for multiple notebook servers.

f. The Mount Path is the location in the container where the volume is accessed. For Data Volumes, the default directory used by Jupyter Notebook is /home/jovyan/-datavol-1.

The second feature is the + Attach Existing Volume button, which is used to attach an existing storage volume to a notebook or machine learning pipeline.

To view the details of the volume attachment, click the expand icon next to the Delete icon. By default, the volume attachment displays the following settings:

a. The volume type is set to Kubernetes volume.

b. You can select Readonly to ensure that the attached volume is mounted in read-only mode, preventing data from being written to or modified.

c. The Name field is used to select an available workspace volume. To create a new volume, refer to the Create Volumes section in this guide.

d. The Mount Path is the location in the container where the volume is accessed. For Workspace Volumes, the default directory used by Jupyter Notebook is /home/jovyan.

Choose GPU Type

To select the GPU type for the notebook server, click Advanced Options.

Select the GPU in the Affinity Config section. Currently, the following GPU options are available:

• NVIDIA HGX-H100 SXM 80GB

• NVIDIA HGX-H100 SXM 10GB (MIG)

• NVIDIA HGX-H100 SXM 40GB (MIG)

• NVIDIA HGX-H100 SXM 20GB (MIG)

• NVIDIA EGX-L40S 48GB

Click the Launch button to create a notebook.

Open JupyterLab Notebook

After creating and running a notebook, click the Connect button to open it. JupyterLab opens automatically.

JupyterLab is the workspace used to create and run notebooks. It provides tools for managing files, editing code, and executing notebook cells through a web-based interface.

Stop JupyterLab Notebook

Stop Notebook Server is used to temporarily stop a running notebook server. It offers several benefits, including resource conservation, preserving the current state, and improved resource management.

A confirmation dialog is displayed before the notebook server is stopped. Review the warning message, then click Stop to continue or Cancel to close the dialog and keep the notebook server running.

Start JupyterLab Notebook

Start Notebook Server is used to start a stopped notebook server and restore access to the notebook environment. After the notebook server is running, you can connect to the notebook and continue your work.

A notification appears to indicate that the notebook server is starting. Once the startup process is complete, the notebook will be available for use.

Delete Notebook

Delete Notebook is used to remove notebooks that are no longer needed. On the Notebooks page, locate the notebook you want to delete, then click the trash icon.

Click the Delete button to continue the notebook deletion process.