flink-ads-processing

A flink + kotlin exercise

Explore

Start the project

docker compose up

this will take a while, go explore flink windowing docs meanwhile.

Once the stack is up, you can

• View jobs on the jobmanager: http://localhost:8081/#/job/running

• Dig into raw msgs and topics on the control-center (kafka ui): http://localhost:9021

• Monitor incoming results on the dashboard: http://localhost:3000/d/fdf835b7-42d2-402a-93b0-3b048a1cfd41/flink-jobs-visualization?orgId=1&refresh=5s

What's in here

flowchart TB
        ws[Web Crawler] --> tsa>topic:\n scraped-ads]
        tsa --- fj1[Flink job:\n AdsByCityYearTumbling]
        tsa --- fj2[Flink job:\n AdsByCitySliding]
        fj1 --> abctw>topic:\n ads-by-city-and-year-tmbl-win]
        fj2 --> abcsw>topic:\n ads-by-city-sld-win]
        abctw --- g[visualization stack]
        abcsw --- g

Loading

Stack Components:

1. A scrapy crawler for collecting ads from a board website
2. A kraft deployment - stores the raw scraped-ads topic, and the output topics from flink aggregation jobs.
3. A Flink cluster + client, to submit and run 2 kotlin aggregation jobs.
4. The Visualization stack - promtail, loki and grafana (honestly I wanted grafana to read directly from kafka but the required plugin relies on librdkafka and won't compile on arm64 😢)

Helpful links (i.e. My code is copied from)

• https://github.com/BarSnir/event_driven_freedom
• https://github.com/katyagorshkova/kafka-kraft
• https://github.com/dfdeshom/scrapy-kafka
• https://github.com/valeriouberti/flink-with-kotlin