Search:

ETSTUR

etstur-image

Challenge

Etstur is the biggest company in Turkey that specialises in Travel & Hotel Booking and Reservation. Etstur handles both sales and marketing of thousands of hotels including over 60 destinations in Turkey and over 40 destinations globally. Etstur also wants to be number one data provider globally by using its 20+ years of experience and to achieve this they needed an alternative system that scales according to their needs. Challenge

Being able to run the same application on every environment has lots of advantages like reducing unexpected bugs or reproducing issues for every environment. To enable this, Etstur decided to dockerize the entire system including Java EE applications, Wildfly application runtime

The next step was to setup a stronger development pipeline to fit into their agile improvements and reduce the deployment complexity to focus on their business logic.

Etstur does several campaigns to provide cheap travel and hotel bookings to their customers, and this causes peaks on their infrastructure. Etstur knows that scalability is the  key concept to reduce their operational costs by scaling up and down according when needed 

For those purposed, on-premise Kubernetes Cluster is provisioned using Ansible in order to minimize the risks during upgrading the cluster.

Client: ETSTUR
Project type: DevOps as a Service
Website: odamax.com

Solution

Before requests were coming to Kubernetes Cluster, there was a HAProxy. In HAProxy, all Kubernetes machines’ IPs were defined. Thus, any request that arrived at HAProxy was redirected to one of the Kubernetes machines.

Whenever it is needed to upgrade the Kubernetes version, the second cluster is provisioned using Ansible and made a session draining using HAProxy. So, basically stack level blue-green is done to minimize the risks. During the upgrade, the traffic is redirected to the new stack.

etstur-results

Results

Etstur are able to upgrade their stack without any risks. If something goes wrong, they can easily switch back to the old system.

Also, they don't redirect the whole traffic to the new stack immediately. It happens in a controlled way.

This gives better results and more confidence.

Contact