More projects.

Centralized VPN setup across On-prem and Cloud.

We helped set up a centralized VPN using OpenVPN for a Fintech client. The client’s infrastructure consisted of 6 data centres across three countries. We set up and managed VPN client-server configuration across on-prem and cloud, enabling the client to connect infrastructure and services across countries and cloud providers securely.

Migrating VM based workloads to Cloud Native stack.

We helped a Fintech client migrate their microservices workloads spread across 40+VM in production to Kubernetes-based deployments with full CI/CD and observability in place. We enabled auto-scaling for stateless API workloads using Kubernetes HPA. This allowed the client to handle peak workloads easily.

Migrating workers from VM to Kubernetes using KEDA.

We implemented Kubernetes Event Driven Autoscaling (KEDA) to migrate background workers running in VMs to Kubernetes. This enabled the client to auto-scale the workers based on load. Earlier, the client team would have to distribute the workers across VMs using a supervisor manually.

Database Reliability use case.

We managed and maintained a highly available master-slave 1TB MySQL 5.7 installation for a Fintech company. We also configured ProxySQL rules to ensure read-write traffic split, load balancing, and rolling upgrades for DB VM instances. We also implemented zero downtime online schema migration for 100GB+ tables with 150M+ rows using GitHub’s gh-ost.

Implementing GitOps for Continuous Deployment.

We moved from the traditional push-based CD pipeline to a pull-based Gitops approach. This prevented sharing credentials across the environments. We used FluxCD for implementation and Git repository as a single source of truth.

Creating fully functional environment per Git branch in Kubernetes.

We created a provisioning framework in Kubernetes that spins up all resources in separate namespaces based on the Git branch. This helped developers to end-to-end tests for their changes in an isolated environment and helped reduce integration errors on production.