Flying with K8s - Cloud Expo Europe
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Load • We are a seasonal business, focussed on the UK market • We see to peaks in January and June. • The site is mostly busy from 9am until 11pm. • We have grown a lot, needed an IT solution that would allow us to continue to grow, without just building for the peak.
IT Approach at Jet2 • We invest heavily in IT. Our development function alone is circa 250 people and will grow to circa 350. • We use a number of 3rd party products where appropriate, but have a large number of development teams in house dedicated to critical business systems, particularly the eCommerce systems.
IT Approach at Jet2 • We are predominately a Microsoft house – we use • This is great & has enabled us to deliver award winning booking and customer service systems.
Problems we were trying to solve • Test • Wanted ability to quickly set up & deploy test environments reliably. • Dynamic test environments with full integration tests • Production • Ability to continue to scale, without continuing to buy peak hardware • Monitoring and alerting in place and consistent • Release Process • Smoother, more automated & repeatable
Technology Selection • Looking for a way to scale & build out a highly dynamic system of constantly changing pricing and availability • Looked at a lot of different technologies & vendors over a 2 year time period • Each previous proof of concepts failed for various reasons, around cost & performance • Wanted to be able to easily scale • Industry consolidated in this time on Docker, Kubernetes, so that was the route we took
Project & Timeline • Initial POC in Azure AKS Feb 2018 – July 2018 • Main project August 2018 – December 2018. • Had a hard deadline to meet for peak trading in January. • Initial focus on flights, rather than holidays, as a lot less complex, without the 3rd party feeds.
Platform Technologies • .Net Core • New event driven system based on Kafka • Kubernetes to manage system • Helm to deploy • Still working on the dev ops pipeline
Kubernetes Selection • Had to deploy on premise to meet initial security requirements. • Didn’t have a lot of time. • Did a brief survey of different vendors, but we’d already seen Rancher at workshops. • It worked for us • Quick to stand up on premise, • Reliable • AD integration • Support • UI for ops
Kubernetes Set Up
Route To Live Cloud On Premise Test On Premise Production • Dynamic Test • Integration • Dark Live Test • Live • Stage
Benefits Gained • Customers have a new experience on our site • https://www.jet2.com/lff • Proved resilience & monitoring • Opportunities for vastly improved continuous delivery to bring change to market faster • Attracting & retaining staff • Surpassed expectations in terms of how fast events could travel through the system. • Demonstrated system could handle peak load with ease
Advice • Two approaches to technology change • Throw everything away and start again • Small, slow, incremental improvements
First Approach Advice • C# only continuity • A lot to deploy, important to focus on what’s required • Get the balance right in terms of speed of delivery vs technical debt • Deliver value early and often • Start simple, build into more complex scenarios • Look to bring in the right external expertise to supplement your existing team • Preproduction / Dark Live is invaluable
Q&A • Any Questions
You can also read