Skip to main content

Serverless API Data Ingestion in Google BigQuery: Part 2(Show Me The Code!)

Ingesting API Data in Google BigQuery the Serverless way!

Introduction

A few days ago I wrote 1st part of this blog. Where I talked about the concept that how can we design serverless data ingestion from API data either as streaming or batch pipeline. In this blog, we will see the code and configuration for the entire pipeline.

Let's get started.

Design

In the previous blog, we discuss pattern 1 where we are ingesting data from API and Inserting it into BigQuery In real-time. We will use this pattern for showing code and configuration.

Cloud Workflow Yaml Configuration

  • Cloud workflow is a serverless offering from GCP that allows us to design workflow that can execute our multistep pipeline as we design. If some steps get failed we can stop or take the necessary steps to alert system operators to know about failure.
  • In our case, we are initializing some variables that we will use later in configurations such as project, datasetId, tableId, region, and SQL query.
  • As per our design, we will execute a cloud function that will ingest data from News API to grab the top headlines.
  • After ingesting data in cloud function we write that data as a streaming insert to our bigquery.
  • Then we execute the SQL query job to compute some metrics and write the result into a result table called “daily_top_headlines_by_source”. we are making a call to the ComputeMetrics sub-workflow & we are also waiting until the job gets finished so that we can say it was successful or unsuccessful.
  • After that, we log out our result of the final step.
      

Cloud Function API Ingestion Logic

  • Let's deep dive into Our Cloud Functions logic. Cloud Functions are serverless functions that allow developers to write lightweight logic and execute in various language/execution environments such as Java, nodejs, python, etc.
  • For our example, I am writing our lightweight logic in Java. I have used HTTP client make an HTTP request to news headline api and grab the response as JSON.
  • Once JSON response is recieved i am using ObjectMapper to read and convert JSON response to JsonNode.
  • JsonNode gives us the flexibility to extract the records and traverse each record and finally grab the important information such as source, author, title, description and so on.
  • We will create a rowContent object that will store map data with key and value.
  • Finally with BigQuery API we will insert row content to target big query table in streaming fashion.

HTTP Client Code

POM file

  • In order to write HTTP client to grab news data and convert JSON to rowContent object and finally write to BigQuery table, we need to add some libraries in our POM.xml file.
  • We have used 2 libraries, google cloud bigquery which provides rowContent Insert method into BigQuery, and Jackson-databind which provides Object mapper and JsonNode to read JSON response and traverse the JSON data.
  • HTTP Client is used from Java 11 SDK and our execution environment support java 11 so no need to add anything in POM.xml.

BigQuery Destination Table

  • Once the cloud function executes successfully we can see the result inserted result in the bigquery table.

Computed Metrics

  • Now as per design 2nd step is to compute the metrics and write it to another table called daily_top_headlines_by_source.
  • We will run the below query and insert the result into our final table “daily_top_headlines_by_source” using bigquery query job that we have configured in cloud workflow.
SELECT
source,count(*) as total_published
FROM
`avid-subject-309421.demo.top_headlines`
GROUP BY
source

After successful execution, we can see that the results have been written to “daily_top_headlines_by_source” table.

Conclusion

In Google Cloud, we can design and build no-ops pipelines very easily with serverless offerings. This blog just shows one of the use cases and barely touches the surfaces. But as your business getting started and you can't afford to have a big engineering team serverless and pay as you go is a savior for your business.

Don't treat Cloud as On-prem and take full advantage of the no-ops data pipeline.

Happy Pipeline Building! ✌️

Comments

Post a Comment

Popular posts from this blog

Beginners Guide to Machine Learning on GCP

This is the title of the webpage! This blog covers basic knowledge needed to get started ML journey on GCP. It provides foundational knowledge which will help readers to gain some level of confidence understanding ML ecosystem on GCP from where they can master each component. Introduction to Machine Learning Machine Learning is a way to use some set of algorithms to derive predictive analytics from data . It is different than Business Intelligence and Data Analytics in a sense that In BI and Data analytics Businesses make decision based on historical data, but In case of Machine Learning , Businesses  predict  the future based on the historical data. Example, It’s a difference between  what happened to the business  vs  what will happen to the business .Its like making BI much smarter and scalable so that it can predict future rather than just showing the state of the business. ML is based on Standard algorithms which are used to create use case specific model
Post a Comment
Read more

Using NPM Library in Google BigQuery UDF

  Javascript UDF’s are cool and using with NPM library is a whole new world to explore! Background One of the main reason to build ETL pipeline was to do data transformation on data before loading into the data warehouse. The only reason we were doing that because data warehouses were not capable to handle these data transformations due to several reasons such as performance and flexibility. In the era of modern data warehouses like Google BigQuery or SnowFlake , things have changed. These data warehouses can process terabyte and petabyte data within seconds and minutes. Considering this much improvement, now performing data transformation within a data warehouse make more sense. Hence to create common transformation logic via UDF (user-defined functions). In this blog, we will see how can we utilize the power of javascript UDF and NPM library to generate data in BigQuery. What is UDF? From Google Cloud Documentations: A user-defined function (UDF) lets you create a function by using
Post a Comment
Read more

What is advertised.listeners in Kafka?

This is the title of the webpage! Hi guys,  Today we gonna talk about Kafka Broker Properties. More Specifically, advertised.listeners property. If you have seen the server.properties file in Kafka there are two properties with listener settings. #listeners=PLAINTEXT://:9092 #advertised.listeners=PLAINTEXT://your.host.name:9092 why the hell we need two listeners for our broker? usually, Kafka brokers talk to each other and register themselves in zookeeper using listeners property. So for all internal cluster communication happens over what you set in listeners property. But if you have a complex network, for example, consider if your cluster is on the cloud which has an internal network and also external IP on which rest of the work can connect to your cluster, in that case, you have to set advertised.listeners property with {EXTERNAL_IP}://{EXTERNAL_PORT}. For Example, If Internal IP is 10.168.4.9 and port is 9092 and External IP is  35.19
3 comments
Read more