Case Study Video with Client
Over the past couple of years I’ve worked with the Australian Associated Press to modernise their legacy environments into Google Cloud. This has involved a large number of implementations and improvements. Their current technology landscape is now vastly improved and due to the customer’s willingness to onboard new processes and ways of working they have seen a great number of benefits.
Their process was a two-staged rocket launch:
- A safe launch into Google Cloud using Google Cloud VMware Engine
- A modernisation out of VMWare Engine into native Google Cloud services.
Below is a video case study that I did with the client to talk about this staging and some of the both business and technical challenges we took on throughout.
Points of Interest
There were so many interesting tools + technologies we utilised during their migration out of legacy and into modern tech. Consider some of these:
- Google Cloud VMware Engine - HCX Layer 2 stretch for zero-downtime migration
- Cloud Storage + FUSE - For enormous photo and video libraries that were accessed in a traditional block storage filesystem manner
- GKE Autopilot - A fully scalable and “nodeless” Kubernetes configuration providing
extremely scalable and agile hosting of some of these core services:
- MongoDB - Using Kubes operators it autoscales Primaries and Arbiters despite node changes
- Postfix is used all their internal mailing systems use a simple HA postfix running in Kubernetes using almost no resources
- Elastic is using the community Elastic resources they are autoscaling and providing full HA for their Elastic, Kibana and Beats. The Kubes operators here are incredible and are fantastic at auto-healing and health management.
- Palo Alto Firewall Appliances - complex routing and advanced NAT management using firewalling appliances in GCP. This allows for their North/South and East/West firewalling and NIC based routing between VPCs and delivers their clientless VPN experiences.
- Migrations + uplift for unsupported MongoDB versions (4.x) using community provided Python migration tooling.
- Bi-directional NAT using TCP/UDP Load Balancers + Palo Alto NGFW
- Cloud Run - Microservices for so, so many applications - including .NET 3.5 tools upgraded to .NET Core.
- Managed Instance Groups - All Windows and Linux VMs that cannot be moved into Cloud Run or GKE have been built out in Managed Instance Groups using Standby instances for auto-healing, auto-scaling and cost optimisation.
- Terraform for Everything - their entire cloud configuration is broken down into various ‘factories’ and delivered as Infrastructure as Code. Providing a central, reviewable, and fully CI/CD integrated location for their whole configuration.
- Atlantis for Terraform - built out and run from Cloud Run of all places. Using Cloud Storage as temp memory store for plans and locks.