How vWork traced its Ruby monolith end to end with Better Stack
vWork makes job scheduling and dispatch software for delivery and field teams. Its engineers send over 1.5 TB of traces and 400 GB of logs a month to Better Stack, which replaced Grafana Cloud and New Relic.
- Industry
- Job scheduling & dispatch software
- Products used
- Logs, Tracing, Metrics, Dashboards, Better Stack MCP
- Switched from
- Grafana Cloud, New Relic
- Customer since
- April 2024
Where they started
vWork makes job scheduling and dispatch software. Its customers' drivers and field workers receive and complete their jobs in vWork's mobile app.
Over the years, vWork's team has run its observability in several setups. It started with a self-hosted ELK stack on AWS and tried variations of it, then moved its logs to self-hosted Grafana and later to Grafana Cloud. Application performance data lived in New Relic. The team kept the separate services connected as well as it could, and getting deeper information out of its logs took a lot of configuration.
Rafael Fonseca, part of a team who looks after vWork's infrastructure, compared the options and introduced the team to Better Stack. Logs moved first, and the work the team had already done to standardize its logs carried straight over. When Rafael saw Better Stack's distributed tracing, he started looking at bringing everything else across too.
“I could see the story of having all of the monitoring data going into one place, and then being able to correlate it.”
How they set it up: tracing a Ruby monolith
The first step was the logs themselves. The team moved from plain JSON lines to rich events, which Better Stack can connect across services, so an engineer can jump from one part of the stack to the next.
Tracing took real engineering. vWork's main application is a large, long-lived Ruby monolith, where off-the-shelf instrumentation didn't always fit, so every new piece of instrumentation has to be checked first. Where nothing fit, Rafael pulled together the parts himself and built a tracing setup that works for the monolith.
Better Stack is built on the OpenTelemetry standard, so instrumentation isn't tied to one vendor's agent. Data from any service, in any language, arrives in the same open format, and AI tools already understand it.
Rafael built much of the tracing with Claude Code connected to the Better Stack MCP. While he instrumented the application, Claude pulled the logs and traces and investigated them in parallel. Once tracing was in place, vWork moved fully to Better Stack and closed its New Relic account. It now sends over 1.5 TB of traces and over 400 GB of logs to Better Stack every month. The team can see into database queries, follow a request between services, and follow a request from the app to the database.
Investigating a sync issue
Ryan King was vWork's mobile developer before he took over leading the engineering team. When a customer reports a syncing issue, his investigation starts in Better Stack.
He looks at the state of the network traffic for that one user at the times they had the problem. Better Stack filters across all of vWork's network calls and returns the results in seconds, which gives him a precise starting point.
On the infrastructure side, Rafael gets the same picture from an alert. When one fires, the team can see the logs, the services that were impacted, and the latency increase on a particular route.
On Grafana, vWork had logs only. Moving between services to follow a request is something the team gained with Better Stack.
“Being able to take that much data about our network calls and filter it in a matter of seconds is pretty remarkable.”
One place instead of several
Now that Ryan leads engineering, he sees the move from both sides: it works better for his developers, and it fits what vWork is trying to achieve as a business. Logs, traces and alerts sit in one platform, so there is one place to look and one account to run.
Investigations have changed too. Rafael tells Claude what he's working on and what isn't happening, and Claude does the legwork across the logs and traces and reports back with what it found. When the team has a question or product feedback, Better Stack's engineers help in a shared Slack channel.
What's next: more of the platform
Ryan is now looking at which other parts of Better Stack vWork could be put to use. Rafael has already connected Better Stack's new Claude plugin, which brings the MCP together with the AI SRE's investigation skills, so the next investigation can start in Claude.
“Being able to plug Better Stack into Claude Code, then develop a feature or investigate an issue and have it pull the logs and traces and investigate all of that in parallel, is a slam dunk.”
Logs, metrics, traces, uptime and incidents in one platform.