Behind GoComet's Air-Gapped Observability with CtrlB
GoComet is transforming how global businesses manage international freight, bringing greater visibility, control, and efficiency to complex logistics operations. Its platform covers a shipment from the first quote to the final invoice. GoProcure runs reverse auctions to get competitive freight rates, GoTrack follows cargo across ocean, air, road and parcel in real time, GoShipment coordinates execution between every stakeholder, and GoInvoice checks freight bills before anyone pays them. More than 230 global brands across 35 countries run their logistics on it, including pharmaceutical manufacturers such as Sun Pharma and Glenmark.
That puts GoComet in the path of live trade. When a carrier update fails to process or a tracking job stalls, a shipper somewhere loses sight of a container, and a delay that could have been flagged early turns into a missed delivery window. Every production incident lands on a customer's supply chain, which is why reliability is non-negotiable for GoComet's engineering team.
The data behind those workflows is sensitive too. Freight rates, invoices and shipment movements for some of the largest manufacturers in the world move through GoComet's services, and the telemetry those services emit carries the same context. For GoComet, where that telemetry lived mattered as much as how fast engineers could search it.
So observability had to run on GoComet's own infrastructure, with no telemetry leaving its network. And it had to cost a fraction of what it did.
This is where CtrlB came in.
The challenge: Full observability without letting telemetry leave the network
GoComet's engineers needed what every production team needs: application and infrastructure logs, distributed traces, dashboards and alerts, all fast enough to use in the middle of an incident. The hard part was where it all had to run. Every byte of telemetry had to stay on-prem, inside a fully air-gapped environment.
That rules out most of the market. SaaS observability platforms are built around shipping data to the vendor's cloud. Assembling a self-hosted stack from several tools brings its own problems: separate stores for logs and traces, schema work every time a service adds a field, and engineers stitching results together by hand while an incident is still open.
Then there was the bill. At the same ingest volume, GoComet's previous observability stack cost roughly 10x what it costs on CtrlB today.
The challenge was to bring logs, traces and the tools that sit on top of them into one place, keep all of it inside GoComet's network, and bring spend down without ingesting less.
The solution: One air-gapped data lake for all of GoComet's telemetry
GoComet now ingests its application and infrastructure telemetry into CtrlB, deployed on-prem and fully air-gapped. Monitoring, debugging and root-cause analysis all run on that single data lake, and nothing leaves GoComet's network.
On top of that lake, GoComet's engineers get the full set of tools they need to run production:
Application and infra logs, with no schema work
Service, host and platform logs land as they are. When a service starts emitting a new field, that field is searchable the moment it arrives. Nobody waits on a mapping change.
Distributed traces, dashboards and alerts
Engineers start from a single log line and follow the full request path across services. Service health, error rates and latency are monitored out of the box.
AI Insight Engine on live telemetry
Error spikes, queue backlogs, latency regressions and rare failure patterns across services surface before anyone writes a query or builds a dashboard.
MCP connectors in the engineering workflow
Engineers and their AI assistants query GoComet's datasets over MCP to debug incidents, validate ingest and answer questions asked in plain language.
None of this needed a second store. Logs, traces and the insights built on them all come from the same lake, so an investigation never has to cross from one tool into another.
The results: Up to 90% lower spend, and every byte stays on-prem
The impact shows up first in the bill, and then in how GoComet's engineers debug.
Telemetry spend: up to 90% lower at the same ingest volume
GoComet ingests the same volume of telemetry it did before and pays roughly a tenth of what it used to. The saving did not come from sending less data or switching anything off.
100% on-prem: no telemetry leaves GoComet's network
CtrlB runs fully air-gapped inside GoComet's own environment. Application logs, infrastructure logs and traces are stored, searched and analyzed there, and the AI Insight Engine and MCP connectors work on that same data without opening a path to the outside.
One lake for application logs, infrastructure logs and traces
Engineers no longer jump between tools to connect a log line to the request that produced it. The pivot from log to trace happens in one place, and the AI Insight Engine flags error spikes and latency regressions on the same data without waiting for someone to ask.