Tracr eliminates blind spots with native-K8s observability and eBPF tracing

About Tracr

Tracr is a blockchain-based platform built to enhance transparency and traceability in the diamond supply chain. Designed to provide verifiable records of a diamond’s journey from mine to market, Tracr leverages blockchain technology to ensure data integrity, security, and provenance verification. 

Tracr's infrastructure is designed with reliability as a core principle, reflecting the critical nature of its role in the diamond supply chain. To ensure high availability and fault tolerance, the team chose AWS and Kubernetes (K8s) as their foundation, enabling seamless deployment and architectures that inherently support resilience. Given that their customers rely on Tracr for an always-available, tamper-proof record of diamond provenance, the team strategically offloads infrastructure complexity by leveraging AWS managed services wherever possible, minimizing operational overhead. As a lean infrastructure team, they avoid managing databases directly, instead relying on AWS’s managed solutions for scalability and reliability.

The Problem: Monitoring blind spots

Running blockchain nodes within Kubernetes presents unique challenges, particularly in maintaining the state and uptime of containers and Persistent Volume Claims (PVCs). Ensuring these underlying resources remain stable is crucial to sustaining a dependable, production-grade blockchain environment.

Tracr was relying on a legacy observability stack based on:

• Elastic Cloud for logs and partial coverage of APM
• Self-hosted Prometheus for infrastructure monitoring

Their existing stack suffered from multiple gaps that affected the team’s visibility into their service:

• Tracing required significant R&D effort: Tracr maintains visibility and quickly detects issues primarily using tracing, which helps them monitor service performance and ensure smooth operations across its distributed architecture. Tracr were using Elastic APM with manual instrumentation inside the code, which led to partial coverage across their services. The SDK-based distributed tracing (DT) approach was not widely implemented due to its complexity, making it difficult to gain full visibility into request flows.‍
• Limited coverage due to cost trade-offs: Tracr had to limit the scope of their observability coverage in order to balance cost. This resulted in blind spots across their infrastructure and application layers, making it more challenging to detect and diagnose issues proactively.‍
• Insufficient Kubernetes visibility: While Tracr deployed an Elastic agent on their nodes to collect infrastructure metrics, it lacked deep Kubernetes insights. It did not provide sufficient visibility into pod health, resource utilization, or K8s-specific issues, leaving engineers without a comprehensive view of their system’s behavior.
  

"A big plus point for us is that it's not just us as DevOps folks who are responsible for observability. It's now the responsibility of the organization. groundcover helped us to get there by democratizing the access to observability insights, making it easy to control and consume.”

- Kristian Lee, DevOps Lead, Tracr

Why groundcover?

• Effortless, auto-instrumented eBPF tracing: Tracr needed deep tracing insights without committing months of R&D resources to instrumenting their stack. groundcover’s eBPF-powered auto-instrumentation allowed them to extract valuable trace data instantly, eliminating the need for extensive manual instrumentation while still gaining full visibility into their application.‍
• Flexibility in adopting OpenTelemetry: With groundcover, Tracr gained powerful eBPF traces without immediately committing to a full OpenTelemetry (OTel) rollout. This provided the flexibility to adopt OTel at their own pace, focusing on areas where it delivers the most value, rather than overhauling their entire stack upfront.‍
• Democratizing deep observability insights: Before groundcover, Tracr’s observability was primarily log-based, limiting the ability to troubleshoot complex issues. Now, tracing is a core part of their workflow, enabling QA, support, and engineering teams to collaborate more effectively. By sharing live trace examples during investigations, teams can diagnose problems faster and reduce resolution times.‍
• A cost-effective BYOC model: Tracr evaluated multiple observability solutions and found that groundcover’s self-hosted BYOC model offered a more predictable and scalable cost structure, compared to solutions that charge based on data volumes. This allowed them to maximize coverage without cost concerns.
• ‍Seamless deployment and Kubernetes-native integration: Deploying groundcover was simple and frictionless, using its Helm chart and native integration with ArgoCD. In contrast to their previous observability stack, which was complex and difficult to manage, groundcover’s smooth deployment process ensured they could quickly realize value without operational headaches.

“The concept for us of being able to get useful tracing data out of the application without adding six months’ worth of work to the development team's backlog was very attractive for everyone”

- Tim Amstrong, Senior DevOps, Tracr

The Impact

With groundcover, Tracr was able to fully migrate from their previous observability stack, replacing their Elastic APM-based solution with a single, unified platform built for Kubernetes-native monitoring. With deep eBPF-based tracing, comprehensive alerting, and seamless K8s integration, the team was now able to:

1. Improve blockchain reliability through deep Kubernetes monitoring: groundcover’s native Kubernetes observability enables Tracr to gain unparalleled visibility into their infrastructure and ensure the reliability of their blockchain operations. Deeper insights into pod health, resource utilization, and container performance, have allowed the team to prevent disruptions to their network by proactively detecting and resolving issues.‍
2. Comprehensive alerting coverage: Using groundcover’s Terraform-managed alerting, the Tracr team has a fully automated and structured alerting system, covering both infrastructure and application-level events. This ensures faster detection of potential failures and significantly reduces operational blind spots.‍
3. Wider adoption across teams: Observability is no longer restricted to a handful of experts. More team members - from DevOps to support and QA - are actively using the platform, thanks to groundcover’s intuitive UI and richer data coverage. Relying on a mix of application logs and limited traces was now a thing of the past, as they now have a unified platform offering full visibility into logs, metrics, and traces.