Migrating to AWS: Delivery & Deployment Uplift for MYOB
MYOB’s in-house DevOps team needed support to achieve their transformation initiatives with automation, tooling and AWS best-practice with minimal onboarding requirements.
DevOps
Uplift
Financial Services
Industry
Business challenge
The MYOB Essentials Production system was hosted in a traditional Australian data centre and due for renewal in early 2016. At the time, MYOB wanted to improve cost efficiencies around 3rd party infrastructure hosting, as well as the time taken to receive additional hosts when needed.
This provided the organisation with an opportunity to innovate and seek out infrastructure improvements to provide delivery teams more control of their environments. The vision was to reduce the cycle time of “concept to production” through gaining configuration traceability, repeatability through automation, and self service to reduce dependency on third parties. This vision ultimately tied into one of MYOB’s core business values to help businesses (their customers) be successful.
“We needed a solution that would enable greater flexibility and visibility in terms of infrastructure configuration, automation and performance”
Grant Tibbey
Infrastructure Lead
Solution
In its current state, the system was running in a third-party data center, utilising SQL Server and a Java front end. The company decided to move its infrastructure to Amazon Web Services (AWS) via CloudFormation to ease maintenance and deploy infrastructure in a repeatable way.
From the time MYOB decided it wanted more flexibility and more customer benefits from the Essentials infrastructure, it had only approximately three months to complete the cutover (until the existing contract expired).
MYOB formed a team using existing in-house DevOps skills but didn’t have enough capacity to meet the aggressive deadlines. MYOB utilises a wide range of technologies and was looking for experience without having to bring additional people up to speed on tools and techniques that were being applied to automation. Cevo were able to offer skills and availability with minimal onboarding requirements.
The existing application had some automation around building and deploying to the application servers using Ansible and Gradle. This coded deployment relied heavily on static infrastructure assets hosted in an unchanging network configuration. By using CloudFormation for network automation, seperate network configurations were created (across 2 availability zones) for elastic load balancers, application and database. A pre-configured VPC was used that was implemented by the core AWS support group and direct-connect was configured via a virtual private interface.
At the time there was no automation around building MS SQL Server, nor around building the network infrastructure. Given the time constraints, existing automation was re-used and augmented with additional tools. Chef was chosen because of its ability to configure Windows and Linux servers securely using Secure WinRM and SSH protocols. It also gave the team the ability to unit and integration test components of the server automation as isolated modules and provided rapid feedback to the development processes using KitchenCI (test-kitchen).
Unlike other automation technologies, code could be developed and tested with the physical server being stubbed. This reduced server development wait-times. CloudFormation was used for network automation and enhanced with complex logic for alarm calculations using ruby code. CloudFormation templates were generated using a ruby gem “cfndsl” to enhance its static nature.
Jenkins was chosen as the CI automation tool as it was previously employed by the Essentials application development team for building, testing and artifact deployment to Nexus Artifact store. The ruby deployment repository was stored in git and Jenkins plugins could be used to create on-demand infrastructure to invoke the AWS component deployment via Cloudformation. The CI toolset decisions took into account the onsite familiarity and knowledge across these deployments platforms.
Options for running MS SQL Server in AWS were considered including RDS, Log Shipping and Always-On Availability Groups to achieve High Availability. These options however were discarded due to cost, extended outage and data loss that would have occurred on failure or reconfiguration. Instead, Mirroring was chosen for HA with Log Shipping initially used to keep the data in AWS in sync with the traditional datacenter to within fifteen minutes so that the cutover would cause minimal impact to customers.
The entire process of building the SQL Cluster in AWS was scripted from scratch. This included formatting, moving, labeling disks, moving system databases and logs, creating users, backup jobs, installing certificates to establishing mirroring using partner and witness amongst many other things that would traditionally have been manually done.
“If something didn’t work, we can tear it down and build it again in hours. Practice also makes perfect. We practiced rebuilding servers, and cutting over the system constantly. Even weeks before the actual cut over took place.”
Grant Tibbey
Infrastructure Lead
By the time cutover took place the entire system could be rebuilt with a couple of commands. Moving forward, the company plans to re-engineer the application from statically deployed to dynamic auto-scaling servers that come and go, and will look at ways to achieve zero downtime deploys to deal with fault tolerance.
With the ongoing use of Chef scripts and development practices applied to Ops, MYOB’s development team can rewrite the application in such a way that it can easily be deployed to AWS. By modularizing and decoupling the deployable application from infrastructure components, they could then be modified, deployed and maintained independently.
Outcomes
There are numerous benefits from implementing the solution; these include providing MYOB with:
- A safe, repeatable and easy means to release features to their customers as soon as they are capable of providing benefit.
- The ability to rebuild servers at will or when they fail.
- The ability to reconfigure or change the way the MYOB Essentials app is built; whether it’s to increase performance, reliability or change how they run deployments, infrastructure architecture can be changed and new services added.
- Cost optimisation – with the scalability offered by AWS, MYOB is able to have full visibility of what is running and can start reducing expenditure through ensuring they are not over or under provisioned.
“[Cevo] collaborated well with our team and were great in every aspect, from planning and delivery, to communications. We were impressed with their continued focus right through to the finish line.”
Thomas Peterson
DevOps Lead
Enjoyed this customer story?
Share it with your network!
You may also like
