Matomo MultiScale

Tekai built a fully automated, production-ready multi-tenant Matomo platform on Kubernetes improving scalability, reducing costs, and enabling near-instant tenant onboarding with high availability.

Matomo MultiScale

Cloud Provider

OVH Public Cloud (EU-WEST-PAR)

IaC

Terraform + OVH Provider

Analytics Application

Matomo

CI/CD

Bitbucket Pipelines

Observability SaaS

Betterstack (logs + metrics)

The image cover was generated by AI

"Me, as a person fully involved in the project from the TrackAd aise was positively surprised with availability of Tekai’s engineering team. I was always able to get my responses in time and, what was especially important, full comprehension on the importance of the project and target of zero downtime policy."

Mikhail Andreiuk, CTO at

TrackAd

About TrackAd

TrackAd is a SaaS company that provides data-driven tools to help businesses analyze and optimize their digital advertising performance across multiple channels. Its platform centralizes marketing data, measures true ROI, and uses AI-powered insights to improve campaign efficiency and decision-making.

The Challenge

The client aims to set up a new Matomo infrastructure on OVH Cloud, migrate more than 10 existing Matomo instances, and implement Bitbucket pipelines to automate instance creation, configuration updates, and future upgrades.

The Solution

  • Team & Approach A Vietnam-based senior DevOps engineer worked closely with TrackAd’s CTO to optimize costs while still delivering strong, reliable results.
  • What We Built The project required to build the entire system from scratch, including setting up cloud infrastructure, designing the Kubernetes architecture, creating reusable multi-tenant deployment pipelines, automating operations, and implementing a full observability stack - all designed to be production-ready from day one.
    • Built a fully automated, production-ready multi-tenant platform on OVH Cloud using Terraform and Kubernetes, with auto-scaling and S3-compatible backups
    • Designed a reusable Helm-based architecture (CQRS) enabling isolated tenants with independent scaling
    • Implemented a pipeline to provision new tenants in under 5 minutes
    • Set up cluster-wide components: ingress, automated TLS, distributed storage (Longhorn), and full observability (logs, metrics, alerts)
    • Developed a custom node auto-recovery mechanism, reducing failure recovery time to minutes
    • Delivered cost-optimized monitoring and a robust disaster recovery setup (RTO ≤ 4h, RPO ≤ 24h)
  • How We Worked Together Throughout the project, our DevOps joined 1 weekly catch-up to report and discuss Q&A with the CTO.

Key Results

  • Node recovery time: reduced from 19 hours to 4–5 minutes (~99.6% faster) with automated self-healing, eliminating manual intervention
  • Observability cost optimization: reduced 80–90% (logs) and 75–85% (metrics) while maintaining full alert coverage
  • Tenant onboarding: improved from 1–2 hours (manual) to under 5 minutes (fully automated)
  • Infrastructure cost per tenant: reduced by ~26% through resource optimization and better capacity planning
  • Platform availability: achieved 99.7%+ uptime (the tracker availability is even higher due to multi-node approach), handling multiple failures with no visible downtime to tenants
Contact us