Skip to content

Latest commit

 

History

History
202 lines (152 loc) · 10.1 KB

README.md

File metadata and controls

202 lines (152 loc) · 10.1 KB
Raw

Talk Dev Fest 2023

What I did

Today I started the Demo that is going to be used on the presentation for the Dev Fest 2023, the title of this presentation is going to be "Lets talk about data and cloud", for this presentation I am going to prepare a Demo with the next stack

  • DBT
  • Docker
  • Airflow
  • AWS S3
  • AWS EC2
  • Snowflake

The propouse of this talk is to show how dbt implements an easy way to configure, build and create a DataWarehouse on Cloud. I was thinking about using Redshift also but it depends.

Today I had the chance to configure and run with success dbt-core on my machine, also configure an Snowflake account and database. Also I had the chance to create a connection between dbt-core with snowflake, and had a chance to execute a dbt run to build some models on Snowflake. To test the connection I execute dbt debug

Questions

How to load this project on Docker?

Fist things first, we need to define a docker file for this kind of task, we can find an example of a dockerFile an the official documentation. So we can base our docker file on it and be able to conteinerize our dbt solution. Don't forget to copy the profiles.yml and hide all the sensitive information in the envfile. https://github.com/dbt-labs/dbt-core/blob/main/docker/Dockerfile

docker run -i -t dbt:latest run

Also we can execute the next comand to know that our connection is working succesfully

docker run -i -t dbt:latest debug

How to load this Docker into AWS?

Firts, I created a new aws account. Lest create the next list of components:

  • awsCluster -> aws-ecs-demo
  • awsTaskDefinition -> aws-task-demo-a:1
  • awsVPC ->
  • awsSG -> aws-sg-demo-a
  • awsEcr -> 466854116461.dkr.ecr.us-east-1.amazonaws.com/aws-repo-dbt-demo

We are going to create a run task to execute this task on the way, we already set all the enviroment variables on the creation of the task. It would be better if we find a way to set an aws kms for those credentials

How to create a security group and allow aws EC2 to run and connecto to Snowflake?

I has been able to execute this ECR and able to reach Snowflake, I had to set a SG and somo vpc changes. aws-sg-demo-a

How to configure AWS Airflow to Run EC2 on demand?

I am dealing with this issue now, I am trying to use AWS MWAA to set an env of airflow easily but some configurations are going wrong on the side of the network. So I am eveluating creating a simple EC2 to set airflow there and then I guess that I will need to deal on HOW to execute thos run task from ec2 or intead we can use another alternative from the aws services such like aws step functions.

Well, at the end I decided to go for StepFunctions, it would be much easier to work with them than defining an airflow env.

{
  "StartAt": "Run Fargate Task",
  "TimeoutSeconds": 3600,
  "States": {
    "Run Fargate Task": {
      "Type": "Task",
      "Resource": "arn:aws:states:::ecs:runTask.sync",
      "Parameters": {
        "LaunchType": "FARGATE",
        "Cluster": "arn:aws:ecs:us-east-1:466854116461:cluster/aws-ecs-demo",
        "TaskDefinition": "arn:aws:ecs:us-east-1:466854116461:task-definition/aws-task-demo-a:1",
        "NetworkConfiguration": {
          "AwsvpcConfiguration": {
            "Subnets": [
              "subnet-0883c91cd73edfbfc",
              "subnet-05a378b1515ec40aa",
              "subnet-00c9ffaa0c22ec432",
              "subnet-07dd11371247c17f7"
            ],
            "AssignPublicIp": "ENABLED"
          }
        }
      },
      "End": true
    }
  }
}

Its funny, maybe I just had not enough sleep but I was dealing with an issue related to roles, but the role parameter is optional, I was dealing with it for a while but now is fixed and the execution of this StateMachine is working.

How to create an AWS Lambda that detect when a file is deposited on the AWS S3, and run the EC2 or instead of using EC2, execute Airflow to execute the pipeline?

There are some templetes that makes the task easy to deploy a lambda function, also in the aws docs you can find how to deploy a lambda function that triggers an aws step functions task. You need also to grant some access to the role that you created to the aws lambda.

import json
import urllib.parse
import boto3

print('Loading function')

s3 = boto3.client('s3')
sf = boto3.client('stepfunctions', region_name = 'us-east-1')

def lambda_handler(event, context):

    try:
        print("SF execution status starting ----> ")
        input_dict = {'key': 'value'}
        response_sf = sf.start_execution(
            stateMachineArn = 'arn:aws:states:us-east-1:466854116461:stateMachine:aws-stepfunc-demo-fargate1',
            input = json.dumps(input_dict)
        )
    
        print("SF execution status --->>>>" + str(response_sf))
        
        return json.dumps(response_sf, default=str)
    except Exception as e:
        print(e)
        raise e

How to read s3 bucket and files from dbt and load them into snowflake?

As part of my research to find a way to have the raw data that is being deposited on the aws s3 bucket, I finded that snowflake provides a way to read the data on aws by creating stage external tables, that allows snowflake to reach the raw data from aws and get it into snowflake.

CREATE STORAGE INTEGRATION s3_int
  TYPE = EXTERNAL_STAGE
  STORAGE_PROVIDER = 'S3'
  ENABLED = TRUE
  STORAGE_AWS_ROLE_ARN = 'arn:aws:iam::466854116461:role/demo-snowflake-role'
  STORAGE_ALLOWED_LOCATIONS = ('s3://bucket-demo-dbt/raw-data/');

DESC INTEGRATION s3_int;

GRANT CREATE STAGE ON SCHEMA public TO ROLE SRJEFERS_READER;
GRANT CREATE FILE FORMAT ON SCHEMA public TO ROLE SRJEFERS_READER;

GRANT USAGE ON INTEGRATION s3_int TO ROLE SRJEFERS_READER;

USE SCHEMA TEST_CONNECTION.public;

CREATE STAGE my_s3_stage3
  STORAGE_INTEGRATION = s3_int
  URL = 's3://bucket-demo-dbt/raw-data/'
  FILE_FORMAT = (TYPE = 'CSV');

LIST @MY_S3_STAGE3;

CREATE TABLE test_schema.test AS 
SELECT 
   -- metadata$filename, 
   -- metadata$file_row_number, 
   t.$1, 
   t.$2
FROM @MY_S3_STAGE3 t;

Access point was also defined poiting to s3.

s3://arn:aws:s3:us-east-1:466854116461:accesspoint/snowflake-vpc
arn:aws:s3:us-east-1:466854116461:accesspoint/snowflake-vpc

As an additional, I am going to add an extra step that is related to the dbt package called dbt-labs/dbt_external_tables to be able to execute and refresh all the external stages that has been created.

Posted a quetion on StackOverflow -> https://stackoverflow.com/questions/77248767/dbt-external-tables-not-being-executed

I saw that maybe those external tables are not supposed to be executed from dbt, on each run, I am going to strar trying to check how those external stages really works and if is requiered to build them on every execution. Also I am starting to look at kaggle for some datasets to use.

I am going to start testing a different approach, it means that snowflake is able to refresh the metadata by itselft without any other external tool, at least for external stages which data source is located in AWS S3.

Solution

I tried to use SQS as I mentioned before it was a great way to fix the pipe executions, it means that I was going to be able to execute them everytime when I had a new/update file on the bucket, since Snowflake provides me a sqs defined by itself from the creation of each PIPE, I started to look for diferent ways to execute it on cascade but since I had no much time to look for another way I decided to look on an old dbt package that I saw that look great to me to implement. dbt-external-tables allows me to execute those tables, from the stages that I defined that are connected via snowflake integration to aws s3, I got the chance to find a way to solve an issue that I got related to the execution, it was very simple since the indent that I got ont the yml was not accurate. Just updating the file gives me the chace to execute it and being able to query those table as sources.

Was great and good enough, now the solution is a dockerized and running on ecs, being executed each time when a file is being uploaded/modified on the s3 bucket, with a lambda function that execute a simple step function.

Next steps

It works on SQS services, I tired to use sns to have an extra service over the sqs but I got many issues when I was trying to configure it, also sqs was a much easier way to set, now I am wondering how to execute this sqs from aws lambda, or another service,

Link to slides

https://docs.google.com/presentation/d/1PvdQbBnsJz5SV43yv9B8yIuWp_u7_MMuqvxWO9aMfAY/edit?usp=sharing

Some links with information