There has been extensive work to port and validate SEBA running on ARM hardware, as described here.

...

Please refer to the SEBA Blueprint Architecture Document for a high level architectural view of the SEBA Blueprint. The same principles and considerations apply here as well.

For a broader, more general view of the SEBA architecture, HW/SW requirements and usecases, please visit the upstream Opencord resources:

• ONF SEBA
• Opencord SEBA Wiki
• CORD Guide - Prereqs

The same priniciples principles and guidelines apply for running SEBA on ARM as well, with the mention that running on ARM equipped HW generally requires a more in-depth knowledge of the HW and SW platform.

...

• Gigabyte™ R120-T32 with Marvell® ThunderX® CN8890 Processors
• Gigabyte™ R281-T91 with Marvell® ThunderX2® CN9975 Processors
• Lenovo™ ThinkSystem HR330A with Ampere eMAG™ HR330A Processors

SEBA for ARM Porting Iteration 1

The initial effort for porting SEBA on ARM started officially in Akraino in early 2019. At that time, the upstream Opencord project had released CORD Platform 6.1 and SEBA 1.0 as described in the CORD 6.1 Guide.

Find more information in the related sub-pages:

• SEBA Porting on Armon ARM - porting and upstreaming work
• SEBA 1.0.0 for ARM Installation instructions

SEBA for ARM Porting Iteration 2

The latest effort for porting SEBA on ARM has since moved away from the old CORD Platform 6.1 and SEBA 1.0 because there was no more real interest in the upstream Opencord project to maintain them, and at the same time the operators were pushing for adding more functionality and make SEBA better in quality and production ready.

...

However, the SEBA 2.0-alpha release was not considered stable enough at the time of the release, so the development work has continued upstream until a more stable, development version was available and running in the Opencord Jenkins CI - SiaB Jobs.

...

could be used as a reference for porting.

Software Platform

During the porting effort, SEBA has been verified almost exclusively using IEC Type 2 Platform. You can read more about the architecture and installation notes from the wiki pages:

• Integrated Edge Cloud (IEC) Blueprint Family
• IEC Type 2 for Integrated Edge Cloud (IEC) Blueprint Family
• IEC Type1&2 Installation Guide for R2

More recently, the work has also continued for verifying SEBA on a TA cluster running on the 3-node Ampere Pod 1 in the Akraino Community Lab at UNH-IOL.

You can get more infomation about the TA Platform from the wiki pages below. Note that TA is most commonly associated with running Radio Edge Cloud (REC) Blueprint, but the TA Platform is common for both REC and SEBA.

• Telco Appliance Blueprint Family
• Porting REC on aarch64 
• REC Installation Guide

Note that at the moment, the work for verifying SEBA on ARM on TA is still in progress.

Base components

For this second iteration, the main base SW components that have been verified are as follows:

Software base

The 2nd interation of the SEBA for ARM port uses a snapshot of the master (development) branch of the various Opencord components, which was taken at a point in time where it was stable enough to be considered for porting.

The components were forked porting was done on several key components from Opencord.

Main helm-charts git repository

Opencord stores all the helm charts components in the helm-charts repository. This allows keeping the charts under version control and tracking changes by the developers.

The porting started out from this repository, which has been forked to the iecedge github account, at some point in the beginning of March 2020. In order to distinguish the iecedge work from the upstream branches, everything is separated on the cord-7.0-arm64 branch, as indicated. The base reference is commit 6622ff3.

iecedge/helm-charts - cord-7.0-arm64

It corresponds roughly to cord-platform-7.2.3 and seba-2.0.0-alpha1 but for specific chart versions of the charts and applications application versions you can consult the following spreadsheets from the Akraino Jira:

SEBA-2.0 for ARM64 - chart versions

In summary, here are the most important components and their chart and application versions:

Additional charts git repository

Besides the CORD related component charts, there are other charts that had to be ported to make SEBA work on ARM architecture.

These components have been divided into two categories and the chart code has been placed in the helm-k8s-charts git repository:

https://github.com/iecedge/helm-k8s-charts

There are two categories, stable and incubator.

The stable category refers to the Kubernetes specific applications and is generally added by default when a Kubernetes cluster is created. 
The incubator category refers to the kafka and zookeeper componentes, where "incubator" is the designation for the helm repository in Opencord parlance.

The table below lists the components by categories:

SEBA-in-a-Box and automation-tools

During the porting phase, it was decided to use SEBA-in-a-Box in the daily work, because it offers better view of the components involved and things that don't work. This has greatly helped with the porting work.
At the same time, SiaB is the only way that Opencord tests SEBA as a whole in daily jobs (see Opencord Jenkins CI - SiaB Jobs), and so it made sense to adopt the same tools and port them to ARM.

The main repository used for SiaB is the Opencord automation-tools. This has been forked to iecedge github account, also under the cord-7.0-arm64 branch. The base reference is commit 5220aeb.

iecedge/automation-tools - cord-7.0-arm64

Test framework - cord-tester

You can read more about the Opencord test framework at these locations:

• CORD Guide - Testing CORD
• System Test Guide
• Automation Framework (CORD)

For the purpose of validating SEBA for ARM in Akraino, only a part of the testing framework has been re-utilized.
Because testing a real usecase requires a complex setup and dedicated PON Hardware, for now only SEBA-in-a-Box has been ported and adapted to work in the Akraino CI.
This means adopting cord-tester and reproducing the SIAB stable Jenkins Job.

The cord-tester was forked to iecedge github account, under cord-7.0-arm64 branch. The base reference commit is 5ce76d4.

iecedge/cord-tester - cord-7.0-arm64

Deliverables and installation notes

Helm repositories

SEBA is generally deployed using 3 set of charts, here referred to as "combined charts" or "release charts":

• cord-platform
• seba
• workflow charts (e.g. att-workflow)

The usual method is to add a helm repository and then install these one by one, as described in the SEBA Blueprint Installation Guide.

For the 2nd iteration of the SEBA porting effort, a helm repository has been created using github.com pages. You can view the helm repo index.yaml here:

https://iecedge.github.io/helm-charts/cord/index.yaml

Additionally, another two helm repositories have been created for storing charts for components not associated with Opencord. This helps to keep the main CORD specific repository free from generic, Kubernetes and third party applications. You can view the indexes of these two helm repositories here:

https://iecedge.github.io/helm-k8s-charts/incubator/index.yaml

https://iecedge.github.io/helm-k8s-charts/stable/index.yaml

Step by step installation instructions as follows:

• Add the CORD repository and updated indexes

helm repo add cord https://iecedge.github.io/helm-charts/cord
helm repo update

• Install the CORD platform

helm install -n cord-platform --version 7.2.3-dev-arm64 cord/cord-platform

• Install SEBA

helm install -n seba --version 2.0.0-dev-arm64 cord/seba

• Install workflow

helm install -n att-workflow --version 1.3.0 cord/att-workflow

The git repositories

It is possible to install SEBA as individual components as well, using the forked iecedge/helm-charts - cord-7.0-arm64.
Note that at this moment, this is the only possible method to install SEBA for ARM by individual components. In contrast, the Opencord also delivers the individual components as packaged helm charts in their https://charts.opencord.org/ helm repo. However, using the git repository has the same effect, so functionality is not lost.

To begin, you need to clone the git repository:

git clone https://github.com/iecedge/helm-charts.git -b cord-7.0-arm64
git clone https://github.com/iecedge/helm-k8s-charts.git -b master

You can follow instructions on the SEBA Installation Guide for installing the individual components. Note that instead of installing the components from the helm repository, you need to install them from the git repository, by specifying the absolute or relative path to the respective directories.

Installing SEBA-in-a-Box

If you decide to install SEBA-in-a-Box for arm64, you need to clone the forked iecedge/automation-tools and run a few commands.

Note that during the porting work, the HW setup implied using iecedge 3-node clusters, so the charts and commands have been tailored to work on 3 nodes, rather than a single node like in upstream Opencord. The same should work just as well on a single node, but it hasn't been verified recently.

If needed, check the information on Integrated Edge Cloud (IEC) Blueprint Family and the IEC Type1&2 Installation Guide for R2.

Also note that automation-tools implies an Ubuntu environment, Ubuntu 18.04 is verified and recommended. Alternatively, you can try running on different distributions, as long as you modify the seba-in-a-box/Makefile accordingly and make sure the necessary dependencies are installed.

git clone https://github.com/iecedge/automation-tools -b cord-7.0-arm64
cd automation-tools/seba-in-a-box
# The env.sh prepares the environment in a way that seba-in-a-box/Makefile expects it to be
# and sets helm-charts to the iecedge specific repo
source env.sh
make stable

Once the installation is finished, you can also try to run the SIAB tests:

make run-tests

Continuous Integration work for SEBA on ARM

As with the other Blueprints in Akraino, SEBA is also periodically verified for correct functionality in the Akraino Jenkins infrastructure.

For the moment, the CI jobs make use of the IEC Type 2 Platform, deployed using the Fuel Installer. Alternatively, the IEC Type 2 Platform can also be deployed using the Compass Installer, which is not in the scope of this documentation.

Hardware setup

Please refer to the Deployment Architecture documentation for better understanding the HW setup and requirements.

At this moment, the Jenkins jobs for deploying and verifying SEBA on ARM, are running on a single POD, located in the ENEA Kista Lab, and connected to the Akraino Jenkins Master.
The jump host that connects to the Akraino Jenkins master is designated by the following Jenkins executor.

https://jenkins.akraino.org/computer/prd-ubuntu1804-dev-96c-256g-1/

There is also work in progress for adding another POD, the ThunderX2 Pod 2 from the Akraino Community Lab at UNH-IOL.

Jenkins pipeline

Currently, SEBA for ARM is verified as part of a bigger pipeline for IEC, which is located under the IEC Jenkins project:

https://jenkins.akraino.org/view/iec/job/iec-type2-fuel-baremetal-ubuntu1804-daily-master/

Note that at this moment, only IEC Type 2 Fuel Baremetal Jenkins jobs support deploying and testing SEBA.
Deploying on the virtual PODs in the Akraino Jenkins brings little value, because of performance penalties.
Also, the Fuel Installer is the only that has been verified, but Compass should work as well. However, there are currently no Compass Baremetal jobs, so the CI needs to be restricted to Fuel baremetal installations for now.

Jenkins jobs

As you can see, the parent job runs a few different jobs:

• iec-type2-deploy-fuel-baremetal-ubuntu1804-daily-master - this deploys the IEC Type 2 Platform using the Fuel Installer and Ubuntu 18.04 as base OS
• validation-enea-daily-master - IEC Blueprint Validation as per Akraino requirements
• iec-type2-install-seba_on_arm-fuel-baremetal-ubuntu1804-daily-master - deploys SEBA-in-a-Box using automation-tools
• iec-type2-test-seba_on_arm-fuel-baremetal-ubuntu1804-daily-master - runs the SEBA-in-a-Box Robot tests

For the test-seba_on_arm job, the robot logs are collected and sent to the Jenkins master for aggregation and visualization.