M30 Milestone Updates

Here are the updates for Milestone M30 (Feb 10 - Feb 23)

• Long-standing bugs were fixed in NeetoCI and NeetoDeploy.

• NeetoDeploy Autoscale gem was causing performance issues in application requests. The gem has been disabled, and the problem has been solved. An improved version of the gem will be released soon.

• Addons were using GP2-type EBS volumes. We switched them to GP3 with optimal IOPS and throughput values, resulting in performance gains and cost reduction.

• Our build machines and review app add-ons were running on AWS spot instances. For more stability, we moved them to on-demand instances. We were able to do this without significant cost difference.

• We completed a POC to set up the Harbor registry in our EKS cluster, with the storage backend as S3. Harbor is a production-grade registry which can be scaled horizontally to support more load in the future. Also, S3 as storage will give us consistent performance, irrespective of the load, at low costs for storage. Image pulls will not incur network costs since we already have an S3 private link setup in our cluster. We have enabled transfer acceleration in S3, which resulted in high-speed image downloads. POC is a success and we will roll this out into production in the next milestone. We expect much faster container starts and re-starts with this change.