Website monitoring / Demo project
FlowWatch
A service that runs key browser workflows and verifies that a lead or order was actually saved on the server.
The challenge
A page can load and a form can report success while the lead itself is lost. The team needs to identify the failing step and confirm when the workflow recovers.
The solution
FlowWatch runs login, lead submission and a test order in Chromium. Persistence is checked through a read-only server endpoint. Two consecutive failures open an incident, and three successful checks confirm recovery.
How it works
The outcome
The report records 39 passing tests, including 12 browser tests. The demo verified one incident, its recovery and two notifications delivered to a local receiver. Tests used the demo site. Live Telegram and Docker runtime were outside this verification.
My contribution
I separated the scheduler, worker and browser process. PostgreSQL tracks state and prevents overlapping runs of the same check. Incidents and notifications are recorded transactionally, and the dashboard shows the failed step and recovery progress.
Architecture
Browser observations go into a shared log that drives incident handling.
Scheduler
Creates scheduled jobs without duplicating slots.
Component connections
- Scheduler: Creates scheduled jobs without duplicating slots.
- Browser scenario: An isolated process executes the workflow and verifies the result on the server.
- Incidents: Ordered observations update the incident and queue a notification.
- Scheduler → Browser scenario
- Browser scenario → Incidents