Subscribe

UiPath AI Center

UiPath AI Center

Configuring the machines

Operating system requirements


🚧

Important!

The machines must run RHEL OS 8.2 or above.

  • We support versions 8.2, 8.3, and 8.4 for any installation. However, if you use a GPU, 8.4 is not supported.

  • Both OS and data disk should be SSD.

  • The OS disk should have at least 512 GiB.

 

Secondary disk requirements


Make sure you allocate a secondary disk for storage with the following requirements:

  • A secondary disk of minimum 512 GiB in size for storage driver. It needs to be mounted and formatted.
  • The default mount location is /datadisk.

 

Disk partitioning


The default partitioning structure on RHEL machines is not suitable for installing Kubernetes. This is because Kubernetes infrastructure is usually installed under the /var partition. By default, the var partition is allocated only 8 GB of space.

For the RHEL OS, you need to ensure you have the following minimum mount point sizes on the machine.

Online

Partition

Online installation

/var

410 GiB

/tmp

20 GiB

/usr

15 GiB

Detailed usage of `/var` folder
  • /var/lib/rancher: 300 GiB (Container Image / layers)

  • /var/lib/kubelet: 100 GiB (Pod / volume mount data)

  • the rest of /var: 10 GiB

Offline

📘

Note:

The requirements for offline are the same as online, except for the first machine where you run the install on, which needs the following requirements.

The extra space is needed to unpack the offline bundle.

Partition

Offline (air-gapped) installation

/var

810 GiB

/tmp

20 GiB

/usr

15 GiB

Detailed usage of `/var` folder
  • /var/lib/rancher: 300 GiB (Container Image / layers)

  • /var/lib/kubelet: 100 GiB (Pod / volume mount data)

  • /var/lib/containers: 150 GiB (Podman local storage)

  • /var/tmp: 250 GiB (for extracting offline bundle)

  • the rest of /var: 10 GiB

 

Enabled ports


Make sure that you have the following ports enabled on your firewall for each source.

Single-node


Port

Protocol

Purpose

22

TCP

For SSH (installation, cluster management debugging)

443

TCP

For HTTPS (accessing AI Center)

Multi-node


Port

Protocol

Purpose

22

TCP

For SSH (installation, cluster management debugging)

443

TCP

For HTTPS (accessing AI Center)

2379

TCP

etcd client port

2380

TCP

etcd peer port

6443

TCP

For accessing Kube API using HTTPS, and required for node joining

8472

UDP

Required for Flannel (VXLAN)

9345

TCP

For accessing Kube API using HTTP, required for node joining

10250

TCP

kubelet / metrics server

30000 - 32767

TCP

NodePort port range for internal communication between nodes in a cluster

🚧

Important!

Ports 443, 6443, and 9345 need to be accessed outside the cluster, but the remaining ports should not be exposed outside the cluster. Run your nodes behind a firewall / security group.

Also ensure, you have connectivity from all nodes to the SQL server, Orchestrator and Identity Server (in case they are different).

If you have a firewall setup in the network, make sure that it has these ports open and allows traffic according to the requirement mentioned above.

 

Configuring GPU nodes


You can provision a GPU on a machine for additional ML workloads. See our guide here: Provisioning a GPU .

Updated 3 months ago

Configuring the machines


Suggested Edits are limited on API Reference Pages

You can only suggest edits to Markdown body content, but not to the API spec.