Sukhraj Singh
All projects

Networking & Java · 2026

Network Support Lab

Find the failed connection. Explain what to check next.

Working local applicationIndependent project

My contribution

I implemented the Java diagnostic path, dependency-aware failure handling and JSON/Markdown support reports.

Evidence

Repository, generated support artifacts, 13 assertions and an uncut 36-second local run.

Scope & limitations

The lab uses controlled targets and remains intentionally read-only; it does not claim production customer access.

  • Java 17
  • Linux
  • TCP/IP
  • PowerShell
  • Python
  • GitHub Actions

See it in action.

00:36 · Local recording
13 assertions -> local customer target -> layered diagnostics -> generated support summary.
About this demonstration

This is an uncut local run: the Java test suite passes 13 assertions, a real local HTTP target starts, interface, DNS, TCP and HTTP checks execute, and the application writes JSON and Markdown handoff artifacts.

13 assertions -> local customer target -> layered diagnostics -> generated support summary.

The problem & approach

The problem

A customer may report that a network management application is unavailable, but that symptom alone does not identify whether the problem begins on the diagnostic host, in DNS, at the network transport layer or inside the application. I wanted a repeatable workflow that collects evidence in dependency order and communicates a bounded next step without overstating the root cause.

The approach

I built a dependency-aware diagnostic tool for a customer-reported network management application problem. It validates the incident input, checks the local interface, DNS, TCP connectivity and HTTP health in order, isolates the first failing layer, and produces both machine-readable evidence and a customer-ready report.

How it works
  1. 01Incident properties
  2. 02Local interface
  3. 03DNS resolution
  4. 04TCP connection
  5. 05HTTP health
  6. 06Assessment + reports
  • I created a validated properties input for the customer, application, target host, service port, health path and timeouts without storing credentials.
  • I implemented ordered Java checks for active local interfaces, hostname resolution, TCP connectivity and the HTTP health response.
  • I marked dependent checks as skipped after a lower-layer failure so one broken dependency does not appear as several unrelated incidents.
  • I classified the first failing layer as local host, DNS, network transport or application, with a healthy baseline when the complete path succeeds.
  • I generated a structured JSON support bundle and a readable Markdown report, then added a Python helper for concise support handoffs.
  • I added PowerShell and Bash build paths, a Linux network-state collector, and GitHub Actions checks for the Java, Python and shell workflow.
  • I verified healthy, HTTP 503, DNS failure, closed-port, skipped-dependency, validation and report-generation scenarios with 13 assertions.
Technical decisions & lessons

Java 17

I used the JVM to model typed configuration, diagnostic results, fault domains and report generation in one dependency-free application.

Layered checks

I ordered the workflow by dependency so DNS must pass before TCP, and TCP must pass before HTTP.

JSON + Markdown

I separated machine-readable evidence from the customer-facing explanation while generating both from the same assessment.

PowerShell, Bash and Python

I used small scripts to keep the workflow practical across Windows and Linux and to create a concise handoff summary.

GitHub Actions

I used CI to compile the Java workflow, run its scenarios and validate the supporting Python and shell files on every push and pull request.

Challenges

  • I had to separate observed evidence from inference so the tool identifies a likely fault domain without claiming a definitive root cause it cannot prove.
  • I had to preserve the dependency chain: an HTTP check should be skipped after a TCP failure rather than reported as a second failure.
  • I kept the lab intentionally read-only; it does not capture packets, use credentials, run remote commands or change customer configuration.

What I learned

  • I learned to troubleshoot from the lowest relevant dependency upward instead of jumping directly to the visible application symptom.
  • I learned how to make Java diagnostics useful to support work by pairing evidence with concise, bounded next actions.
  • I learned why a known-good healthy run is valuable when comparing intermittent customer incidents and platform logs.
  • I learned to treat automated classification as a support aid, not a replacement for customer timelines, routing evidence or application logs.

Dependency-aware result · diagnostic

LOCAL_INTERFACE  PASS
DNS              PASS
TCP              FAIL
HTTP             SKIPPED

faultDomain=NETWORK_TRANSPORT