Deploy Edgeworker Nodes

Introduction

Edgeworker is a new node personality introduced in StarlingX R5.0 [[1]]. Edgeworkers are typically small systems running dedicated workloads that can now be managed by StarlingX. To use the edgeworker personality, nodes must have the Ubuntu operating system installed [[2]].

The edgeworker personality is targeted for nodes that meet one or more of the following criteria:

  • A node that is running a customized OS, or

  • A node that is running a Type1 hypervisor, or

  • A node that does not meet StarlingX worker’s minimum requirements.

These nodes can be added to a running StarlingX cluster in minutes. During the provisioning stage, Kubernetes components will be provisioned on the edgeworker nodes.

The StarlingX controller manages edgeworker nodes using the same mechanisms as typical worker nodes, including the following:

  • Container orchestration by Kubernetes

  • Distributed storage cluster management by Rook-ceph

Notes

Hardware Requirements

The minimum requirements for an edgeworker host (bare metal or virtual machine) are:

Minimum Requirement

Edgeworker Node

Minimum processor class

Intel® Core™ Processor Family

Minimum memory

16GB

Primary disk

256GB SSD or NVMe

Minimum network ports

  • Mgmt/Cluster: 1 x 1000BASE

  • Data: 0 or more x 10GE

Prepare Edgeworker Nodes

Before you provision edgeworker nodes, you must prepare the following setup:

  • Deploy StarlingX AIO Duplex or Standard setup.

  • Connect the Edgeworker node to power and power it on.

  • Connect the Edgeworker node to the network and connect it to a management network switch.

  • Set the primary disk as the first boot device in the BIOS setting of the edgeworker node.

  • Install Ubuntu 18.04 (or above) on the edgeworker node.

  • Install Openssh-server and Python.

The figure below shows how edgeworker nodes are connected to an existing StarlingX cluster.

../../_images/edgeworker-deployment.png

Provision Edgeworker Nodes

When a node’s network interface is connected to the management network switch, the node is detected and will appear in the list of hosts managed by system inventory (with personality set to None).

[sysadmin@controller-0 ~(keystone_admin)]$ system host-list
+----+--------------+-------------+----------------+-------------+--------------+
| id | hostname     | personality | administrative | operational | availability |
+----+--------------+-------------+----------------+-------------+--------------+
| 1  | controller-0 | controller  | unlocked       | enabled     | available    |
| 2  | controller-1 | controller  | unlocked       | enabled     | available    |
| 3  | None         | None        | locked         | disabled    | offline      |
+----+--------------+-------------+----------------+-------------+--------------+

  1. Update the node’s hostname and personality using system host-update:

    [sysadmin@controller-0 ~(keystone_admin)]$ system host-update 3 hostname=edgeworker-0 personality=edgeworker
+-----------------------+--------------------------------------+
| Property              | Value                                |
+-----------------------+--------------------------------------+
| action                | none                                 |
| administrative        | locked                               |
| availability          | offline                              |
| bm_ip                 | None                                 |
| bm_type               | None                                 |
| bm_username           | None                                 |
| boot_device           | /dev/sda                             |
| capabilities          | {}                                   |
| clock_synchronization | ntp                                  |
| config_applied        | None                                 |
| config_status         | None                                 |
| config_target         | None                                 |
| console               | ttyS0,115200                         |
| created_at            | 2021-02-05T02:12:42.136121+00:00     |
| device_image_update   | None                                 |
| hostname              | edgeworker-0                         |
| id                    | 3                                    |
| install_output        | text                                 |
| install_state         | None                                 |
| install_state_info    | None                                 |
| inv_state             | None                                 |
| invprovision          | None                                 |
| location              | {}                                   |
| mgmt_ip               | 192.188.204.21                       |
| mgmt_mac              | 52:54:00:76:e6:cb                    |
| operational           | disabled                             |
| personality           | edgeworker                           |
| reboot_needed         | False                                |
| reserved              | False                                |
| rootfs_device         | /dev/sda                             |
| serialid              | None                                 |
| software_load         | nn.nn                                |
| task                  | None                                 |
| tboot                 | false                                |
| ttys_dcd              | None                                 |
| updated_at            | None                                 |
| uptime                | 0                                    |
| uuid                  | a2c8f3ec-1dc4-4a0a-b471-6e5c01abc187 |
| vim_progress_status   | None                                 |
+-----------------------+--------------------------------------+

    Alternatively, if the node is not shown in the host list automatically, you can also add the host using system host-add:

    [sysadmin@controller-0 ~(keystone_admin)]$ system host-add -n edgeworker-0 -p edgeworker -m <mgmt_mac>

  2. Check the prerequisites of the edgeworker node:

    Note

    We suggest using the same username and password for all the edgeworker nodes.

    EWUSER=<edgeworker username>
EWPASSWD=<edgeworker password>
SYSADMINPASSWD=<sysadmin password>
EWNODE=(edgeworker-0)
PYTHONBIN=()
for NODE in ${EWNODE[@]}; do
  EDGEWORKER=$(system host-show ${NODE} | awk '/mgmt_ip/{print$4}');
  PYTHONBIN+=($(sshpass -p ${EWPASSWD} ssh -o "StrictHostKeyChecking no" ${EWUSER}@${EDGEWORKER} which python));
done
ERR=0
for i in ${PYTHONBIN[@]}; do
  if [ -z "$i" ]; then
    ERR=1
    break
  fi
done
if [ $ERR == 1 ]; then
  echo "Error, python should be intalled on edgeworker nodes."
else
  echo "Pre-requisites check passed."
fi

    If the following error occurs, refresh the DHCP client for the management interface on the edgeworker nodes to assign the right IP address.

    ssh: connect to host XX.XX.XX.XX port 22: No route to host

  3. Create an edgeworker inventory file with the variables from the previous step.

    tee ./edgeworker_inventory.yml << EOF
all:
  hosts:
    localhost:
      ansible_connection: local
  children:
    edgeworker:
      hosts:
EOF
for i in ${!EWNODE[@]}; do
  echo "        ${EWNODE[$i]}:"
  echo "          ansible_ssh_user: ${EWUSER}"
  echo "          ansible_ssh_pass: ${EWPASSWD}"
  echo "          ansible_become_pass: ${EWPASSWD}"
  echo "          ansible_python_interpreter: ${PYTHONBIN[$i]}"
done | tee -a edgeworker_inventory.yml
tee -a ./edgeworker_inventory.yml << EOF
  vars:
    ansible_ssh_user: sysadmin
    ansible_ssh_pass: ${SYSADMINPASSWD}
    ansible_become_pass: ${SYSADMINPASSWD}
EOF

  4. Provision the edgeworker node using Ansible playbook. You can provision edgeworker nodes one at a time or multiple nodes at once. Run the playbook with the inventory file you created.

    ansible-playbook -i ./edgeworker_inventory.yml /usr/share/ansible/stx-ansible/playbooks/provision_edgeworker.yml

    Provisioning output is similar to the following:

    TASK [provision-edgeworker/prepare-edgeworker/kubernetes : set_fact] *********************************************************************************************************
ok: [edgeworker-0]

TASK [provision-edgeworker/prepare-edgeworker/kubernetes : Pull k8s gcr images from controller registry to edgeworker-0] *****************************************************
changed: [edgeworker-0]

RUNNING HANDLER [provision-edgeworker/prepare-edgeworker/kubernetes : Restart kubelet] ***************************************************************************************
changed: [edgeworker-0]

PLAY [localhost] *************************************************************************************************************************************************************

TASK [provision-edgeworker/prepare-controller/cleanup : Cleanup temporary files] *********************************************************************************************
changed: [localhost]

TASK [provision-edgeworker/prepare-controller/cleanup : Remove the provision flag] *******************************************************************************************
changed: [localhost]

PLAY RECAP *******************************************************************************************************************************************************************
edgeworker-0               : ok=54   changed=10   unreachable=0    failed=0
localhost                  : ok=31   changed=19   unreachable=0    failed=0

    After provisioning, the edgeworker node will show Ready status in the Kubernetes cluster.

    [sysadmin@controller-0 ~(keystone_admin)]$ kubectl get node
NAME           STATUS   ROLES    AGE   VERSION
controller-0   Ready    master   42h   v1.18.1
controller-1   Ready    master   42h   v1.18.1
edgeworker-0   Ready    <none>   21m   v1.18.1

Note

In StarlingX R5.0, the edgeworker nodes will be shown as locked/disabled/offline when you use the system host-list command.