Megh Vyas
Backend · Infra · SRE
← All work

SRE monitoring

HealthTrack

A single pane of glass for service health, with the SRE math done for you and streamed live.

0

services, one command

P95 / P99

plus Apdex, computed

Kafka

live event stream

Up is not healthy

Knowing a service is up is easy. Knowing whether it is healthy, its tail latency, its error budget, its Apdex, at a glance, across every dependency, usually means stitching five tools together and still guessing.

One pane of glass

An SRE dashboard where every signal that matters sits on one screen, history is queryable, and alerts fire on the composite picture instead of a single flapping check.

The instrumentation

01

Built configurable HTTP health checks that compute the real SRE numbers: P95 and P99 latency, Apdex, error rate, and uptime, not just a green dot.

02

Streamed every check result to Kafka topics, so other systems can react to health as an event instead of polling for it.

03

Backed multi-signal alerting (status plus Apdex plus error rate) with Redis-based deduplication, so one outage is one page, not a hundred.

04

Stored time series in Prometheus and embedded Grafana dashboards directly in the UI for deep dives.

05

Packaged the entire nine-service stack behind a single Docker command, TypeScript on the front, Python and Pydantic on the back.

On the board

  • P95 / P99 latency, Apdex, error rate, uptime

  • Kafka event streaming of health results

  • Multi-signal alerting with Redis dedup

  • Prometheus time series + embedded Grafana

  • TypeScript frontend, Python 3.11 + Pydantic backend

  • Nine-service stack, one Docker command