Friday, November 22, 2024

Orchestrate an end-to-end ETL pipeline utilizing Amazon S3, AWS Glue, and Amazon Redshift Serverless with Amazon MWAA

Amazon Managed Workflows for Apache Airflow (Amazon MWAA) is a managed orchestration service for Apache Airflow that you need to use to arrange and function information pipelines within the cloud at scale. Apache Airflow is an open supply instrument used to programmatically creator, schedule, and monitor sequences of processes and duties, known as workflows. With Amazon MWAA, you need to use Apache Airflow and Python to create workflows with out having to handle the underlying infrastructure for scalability, availability, and safety.

Through the use of a number of AWS accounts, organizations can successfully scale their workloads and handle their complexity as they develop. This method offers a sturdy mechanism to mitigate the potential influence of disruptions or failures, ensuring that essential workloads stay operational. Moreover, it allows value optimization by aligning sources with particular use instances, ensuring that bills are nicely managed. By isolating workloads with particular safety necessities or compliance wants, organizations can keep the very best ranges of knowledge privateness and safety. Moreover, the power to prepare a number of AWS accounts in a structured method means that you can align your corporation processes and sources in line with your distinctive operational, regulatory, and budgetary necessities. This method promotes effectivity, flexibility, and scalability, enabling giant enterprises to satisfy their evolving wants and obtain their objectives.

This publish demonstrates the way to orchestrate an end-to-end extract, rework, and cargo (ETL) pipeline utilizing Amazon Easy Storage Service (Amazon S3), AWS Glue, and Amazon Redshift Serverless with Amazon MWAA.

Answer overview

For this publish, we take into account a use case the place a knowledge engineering group desires to construct an ETL course of and provides one of the best expertise to their end-users once they wish to question the newest information after new uncooked information are added to Amazon S3 within the central account (Account A within the following structure diagram). The information engineering group desires to separate the uncooked information into its personal AWS account (Account B within the diagram) for elevated safety and management. Additionally they wish to carry out the info processing and transformation work in their very own account (Account B) to compartmentalize duties and stop any unintended modifications to the supply uncooked information current within the central account (Account A). This method permits the group to course of the uncooked information extracted from Account A to Account B, which is devoted for information dealing with duties. This makes certain the uncooked and processed information will be maintained securely separated throughout a number of accounts, if required, for enhanced information governance and safety.

Our resolution makes use of an end-to-end ETL pipeline orchestrated by Amazon MWAA that appears for brand spanking new incremental information in an Amazon S3 location in Account A, the place the uncooked information is current. That is finished by invoking AWS Glue ETL jobs and writing to information objects in a Redshift Serverless cluster in Account B. The pipeline then begins operating saved procedures and SQL instructions on Redshift Serverless. Because the queries end operating, an UNLOAD operation is invoked from the Redshift information warehouse to the S3 bucket in Account A.

As a result of safety is necessary, this publish additionally covers the way to configure an Airflow connection utilizing AWS Secrets and techniques Supervisor to keep away from storing database credentials inside Airflow connections and variables.

The next diagram illustrates the architectural overview of the parts concerned within the orchestration of the workflow.

The workflow consists of the next parts:

  • The supply and goal S3 buckets are in a central account (Account A), whereas Amazon MWAA, AWS Glue, and Amazon Redshift are in a special account (Account B). Cross-account entry has been arrange between S3 buckets in Account A with sources in Account B to have the ability to load and unload information.
  • Within the second account, Amazon MWAA is hosted in a single VPC and Redshift Serverless in a special VPC, that are linked via VPC peering. A Redshift Serverless workgroup is secured inside personal subnets throughout three Availability Zones.
  • Secrets and techniques like person identify, password, DB port, and AWS Area for Redshift Serverless are saved in Secrets and techniques Supervisor.
  • VPC endpoints are created for Amazon S3 and Secrets and techniques Supervisor to work together with different sources.
  • Normally, information engineers create an Airflow Directed Acyclic Graph (DAG) and commit their modifications to GitHub. With GitHub actions, they’re deployed to an S3 bucket in Account B (for this publish, we add the information into S3 bucket instantly). The S3 bucket shops Airflow-related information like DAG information, necessities.txt information, and plugins. AWS Glue ETL scripts and property are saved in one other S3 bucket. This separation helps keep group and keep away from confusion.
  • The Airflow DAG makes use of varied operators, sensors, connections, duties, and guidelines to run the info pipeline as wanted.
  • The Airflow logs are logged in Amazon CloudWatch, and alerts will be configured for monitoring duties. For extra info, see Monitoring dashboards and alarms on Amazon MWAA.

Stipulations

As a result of this resolution facilities round utilizing Amazon MWAA to orchestrate the ETL pipeline, that you must arrange sure foundational sources throughout accounts beforehand. Particularly, that you must create the S3 buckets and folders, AWS Glue sources, and Redshift Serverless sources of their respective accounts previous to implementing the complete workflow integration utilizing Amazon MWAA.

Deploy sources in Account A utilizing AWS CloudFormation

In Account A, launch the supplied AWS CloudFormation stack to create the next sources:

  • The supply and goal S3 buckets and folders. As a finest observe, the enter and output bucket constructions are formatted with hive model partitioning as s3://<bucket>/merchandise/YYYY/MM/DD/.
  • A pattern dataset referred to as merchandise.csv, which we use on this publish.

Add the AWS Glue job to Amazon S3 in Account B

In Account B, create an Amazon S3 location referred to as aws-glue-assets-<account-id>-<area>/scripts (if not current). Exchange the parameters for the account ID and Area within the sample_glue_job.py script and add the AWS Glue job file to the Amazon S3 location.

Deploy sources in Account B utilizing AWS CloudFormation

In Account B, launch the supplied CloudFormation stack template to create the next sources:

  • The S3 bucket airflow-<username>-bucket to retailer Airflow-related information with the next construction:
    • dags – The folder for DAG information.
    • plugins – The file for any customized or neighborhood Airflow plugins.
    • necessities – The necessities.txt file for any Python packages.
    • scripts – Any SQL scripts used within the DAG.
    • information – Any datasets used within the DAG.
  • A Redshift Serverless atmosphere. The identify of the workgroup and namespace are prefixed with pattern.
  • An AWS Glue atmosphere, which comprises the next:
    • An AWS Glue crawler, which crawls the info from the S3 supply bucket sample-inp-bucket-etl-<username> in Account A.
    • A database referred to as products_db within the AWS Glue Information Catalog.
    • An ELT job referred to as sample_glue_job. This job can learn information from the merchandise desk within the Information Catalog and cargo information into the Redshift desk merchandise.
  • A VPC gateway endpointto Amazon S3.
  • An Amazon MWAA atmosphere. For detailed steps to create an Amazon MWAA atmosphere utilizing the Amazon MWAA console, check with Introducing Amazon Managed Workflows for Apache Airflow (MWAA).

launch stack 1

Create Amazon Redshift sources

Create two tables and a saved process on an Redshift Serverless workgroup utilizing the merchandise.sql file.

On this instance, we create two tables referred to as merchandise and products_f. The identify of the saved process is sp_products.

Configure Airflow permissions

After the Amazon MWAA atmosphere is created efficiently, the standing will present as Accessible. Select Open Airflow UI to view the Airflow UI. DAGs are routinely synced from the S3 bucket and visual within the UI. Nonetheless, at this stage, there are not any DAGs within the S3 folder.

Add the shopper managed coverage AmazonMWAAFullConsoleAccess, which grants Airflow customers permissions to entry AWS Id and Entry Administration (IAM) sources, and connect this coverage to the Amazon MWAA position. For extra info, see Accessing an Amazon MWAA atmosphere.

The insurance policies connected to the Amazon MWAA position have full entry and should solely be used for testing functions in a safe check atmosphere. For manufacturing deployments, comply with the least privilege precept.

Arrange the atmosphere

This part outlines the steps to configure the atmosphere. The method entails the next high-level steps:

  1. Replace any crucial suppliers.
  2. Arrange cross-account entry.
  3. Set up a VPC peering connection between the Amazon MWAA VPC and Amazon Redshift VPC.
  4. Configure Secrets and techniques Supervisor to combine with Amazon MWAA.
  5. Outline Airflow connections.

Replace the suppliers

Comply with the steps on this part in case your model of Amazon MWAA is lower than 2.8.1 (the newest model as of scripting this publish).

Suppliers are packages which are maintained by the neighborhood and embody all of the core operators, hooks, and sensors for a given service. The Amazon supplier is used to work together with AWS companies like Amazon S3, Amazon Redshift Serverless, AWS Glue, and extra. There are over 200 modules throughout the Amazon supplier.

Though the model of Airflow supported in Amazon MWAA is 2.6.3, which comes bundled with the Amazon supplied package deal model 8.2.0, help for Amazon Redshift Serverless was not added till the Amazon supplied package deal model 8.4.0. As a result of the default bundled supplier model is older than when Redshift Serverless help was launched, the supplier model should be upgraded as a way to use that performance.

Step one is to replace the constraints file and necessities.txt file with the right variations. Consult with Specifying newer supplier packages for steps to replace the Amazon supplier package deal.

  1. Specify the necessities as follows:
    --constraint "/usr/native/airflow/dags/constraints-3.10-mod.txt"
    apache-airflow-providers-amazon==8.4.0

  2. Replace the model within the constraints file to eight.4.0 or larger.
  3. Add the constraints-3.11-updated.txt file to the /dags folder.

Consult with Apache Airflow variations on Amazon Managed Workflows for Apache Airflow for proper variations of the constraints file relying on the Airflow model.

  1. Navigate to the Amazon MWAA atmosphere and select Edit.
  2. Beneath DAG code in Amazon S3, for Necessities file, select the newest model.
  3. Select Save.

This can replace the atmosphere and new suppliers can be in impact.

  1. To confirm the suppliers model, go to Suppliers below the Admin desk.

The model for the Amazon supplier package deal needs to be 8.4.0, as proven within the following screenshot. If not, there was an error whereas loading necessities.txt. To debug any errors, go to the CloudWatch console and open the requirements_install_ip log in Log streams, the place errors are listed. Consult with Enabling logs on the Amazon MWAA console for extra particulars.

Arrange cross-account entry

That you must arrange cross-account insurance policies and roles between Account A and Account B to entry the S3 buckets to load and unload information. Full the next steps:

  1. In Account A, configure the bucket coverage for bucket sample-inp-bucket-etl-<username> to grant permissions to the AWS Glue and Amazon MWAA roles in Account B for objects in bucket sample-inp-bucket-etl-<username>:
    {
        "Model": "2012-10-17",
        "Assertion": [
            {
                "Effect": "Allow",
                "Principal": {
                    "AWS": [
                        "arn:aws:iam::<account-id-of- AcctB>:role/service-role/<Glue-role>",
                        "arn:aws:iam::<account-id-of-AcctB>:role/service-role/<MWAA-role>"
                    ]
                },
                "Motion": [
                    "s3:GetObject",
    "s3:PutObject",
    		   "s3:PutObjectAcl",
    		   "s3:ListBucket"
                ],
                "Useful resource": [
                    "arn:aws:s3:::sample-inp-bucket-etl-<username>/*",
                    "arn:aws:s3:::sample-inp-bucket-etl-<username>"
                ]
            }
        ]
    }
    

  2. Equally, configure the bucket coverage for bucket sample-opt-bucket-etl-<username> to grant permissions to Amazon MWAA roles in Account B to place objects on this bucket:
    {
        "Model": "2012-10-17",
        "Assertion": [
            {
                "Effect": "Allow",
                "Principal": {
                    "AWS": "arn:aws:iam::<account-id-of-AcctB>:role/service-role/<MWAA-role>"
                },
                "Action": [
                    "s3:GetObject",
                    "s3:PutObject",
                    "s3:PutObjectAcl",
                    "s3:ListBucket"
                ],
                "Useful resource": [
                    "arn:aws:s3:::sample-opt-bucket-etl-<username>/*",
                    "arn:aws:s3:::sample-opt-bucket-etl-<username>"
                ]
            }
        ]
    }
    

  3. In Account A, create an IAM coverage referred to as policy_for_roleA, which permits crucial Amazon S3 actions on the output bucket:
    {
        "Model": "2012-10-17",
        "Assertion": [
            {
                "Sid": "VisualEditor0",
                "Effect": "Allow",
                "Action": [
                    "kms:Decrypt",
                    "kms:Encrypt",
                    "kms:GenerateDataKey"
                ],
                "Useful resource": [
                    "<KMS_KEY_ARN_Used_for_S3_encryption>"
                ]
            },
            {
                "Sid": "VisualEditor1",
                "Impact": "Enable",
                "Motion": [
                    "s3:PutObject",
                    "s3:GetObject",
                    "s3:GetBucketAcl",
                    "s3:GetBucketCors",
                    "s3:GetEncryptionConfiguration",
                    "s3:GetBucketLocation",
                    "s3:ListAllMyBuckets",
                    "s3:ListBucket",
                    "s3:ListBucketMultipartUploads",
                    "s3:ListBucketVersions",
                    "s3:ListMultipartUploadParts"
                ],
                "Useful resource": [
                    "arn:aws:s3:::sample-opt-bucket-etl-<username>",
                    "arn:aws:s3:::sample-opt-bucket-etl-<username>/*"
                ]
            }
        ]
    }

  4. Create a brand new IAM position referred to as RoleA with Account B because the trusted entity position and add this coverage to the position. This enables Account B to imagine RoleA to carry out crucial Amazon S3 actions on the output bucket.
  5. In Account B, create an IAM coverage referred to as s3-cross-account-access with permission to entry objects within the bucket sample-inp-bucket-etl-<username>, which is in Account A.
  6. Add this coverage to the AWS Glue position and Amazon MWAA position:
    {
        "Model": "2012-10-17",
        "Assertion": [
            {
                "Effect": "Allow",
                "Action": [
                    "s3:GetObject",
                    "s3:PutObject",
                    "s3:PutObjectAcl"
                ],
                "Useful resource": "arn:aws:s3:::sample-inp-bucket-etl-<username>/*"
            }
        ]
    }

  7. In Account B, create the IAM coverage policy_for_roleB specifying Account A as a trusted entity. The next is the belief coverage to imagine RoleA in Account A:
    {
        "Model": "2012-10-17",
        "Assertion": [
            {
                "Sid": "CrossAccountPolicy",
                "Effect": "Allow",
                "Action": "sts:AssumeRole",
                "Resource": "arn:aws:iam::<account-id-of-AcctA>:role/RoleA"
            }
        ]
    }

  8. Create a brand new IAM position referred to as RoleB with Amazon Redshift because the trusted entity sort and add this coverage to the position. This enables RoleB to imagine RoleA in Account A and in addition to be assumable by Amazon Redshift.
  9. Connect RoleB to the Redshift Serverless namespace, so Amazon Redshift can write objects to the S3 output bucket in Account A.
  10. Connect the coverage policy_for_roleB to the Amazon MWAA position, which permits Amazon MWAA to entry the output bucket in Account A.

Consult with How do I present cross-account entry to things which are in Amazon S3 buckets? for extra particulars on organising cross-account entry to things in Amazon S3 from AWS Glue and Amazon MWAA. Consult with How do I COPY or UNLOAD information from Amazon Redshift to an Amazon S3 bucket in one other account? for extra particulars on organising roles to unload information from Amazon Redshift to Amazon S3 from Amazon MWAA.

Arrange VPC peering between the Amazon MWAA and Amazon Redshift VPCs

As a result of Amazon MWAA and Amazon Redshift are in two separate VPCs, that you must arrange VPC peering between them. You should add a path to the route tables related to the subnets for each companies. Consult with Work with VPC peering connections for particulars on VPC peering.

Guarantee that CIDR vary of the Amazon MWAA VPC is allowed within the Redshift safety group and the CIDR vary of the Amazon Redshift VPC is allowed within the Amazon MWAA safety group, as proven within the following screenshot.

If any of the previous steps are configured incorrectly, you might be more likely to encounter a “Connection Timeout” error within the DAG run.

Configure the Amazon MWAA reference to Secrets and techniques Supervisor

When the Amazon MWAA pipeline is configured to make use of Secrets and techniques Supervisor, it should first search for connections and variables in an alternate backend (like Secrets and techniques Supervisor). If the alternate backend comprises the wanted worth, it’s returned. In any other case, it should examine the metadata database for the worth and return that as an alternative. For extra particulars, check with Configuring an Apache Airflow connection utilizing an AWS Secrets and techniques Supervisor secret.

Full the next steps:

  1. Configure a VPC endpoint to hyperlink Amazon MWAA and Secrets and techniques Supervisor (com.amazonaws.us-east-1.secretsmanager).

This enables Amazon MWAA to entry credentials saved in Secrets and techniques Supervisor.

  1. To supply Amazon MWAA with permission to entry Secrets and techniques Supervisor secret keys, add the coverage referred to as SecretsManagerReadWrite to the IAM position of the atmosphere.
  2. To create the Secrets and techniques Supervisor backend as an Apache Airflow configuration choice, go to the Airflow configuration choices, add the next key-value pairs, and save your settings.

This configures Airflow to search for connection strings and variables on the airflow/connections/* and airflow/variables/* paths:

secrets and techniques.backend: airflow.suppliers.amazon.aws.secrets and techniques.secrets_manager.SecretsManagerBackend secrets and techniques.backend_kwargs: {"connections_prefix" : "airflow/connections", "variables_prefix" : "airflow/variables"}

  1. To generate an Airflow connection URI string, go to AWS CloudShell and enter right into a Python shell.
  2. Run the next code to generate the connection URI string:
    import urllib.parse
    conn_type="redshift"
    host="sample-workgroup.<account-id-of-AcctB>.us-east-1.redshift-serverless.amazonaws.com" #Specify the Amazon Redshift workgroup endpoint
    port="5439"
    login = 'admin' #Specify the username to make use of for authentication with Amazon Redshift
    password = '<password>' #Specify the password to make use of for authentication with Amazon Redshift
    role_arn = urllib.parse.quote_plus('arn:aws:iam::<account_id>:position/service-role/<MWAA-role>')
    database="dev"
    area = 'us-east-1' #YOUR_REGION
    conn_string = '{0}://{1}:{2}@{3}:{4}?role_arn={5}&database={6}&area={7}'.format(conn_type, login, password, host, port, role_arn, database, area)
    print(conn_string)
    

The connection string needs to be generated as follows:

redshift://admin:<password>@sample-workgroup.<account_id>.us-east-1.redshift-serverless.amazonaws.com:5439?role_arn=<MWAA position ARN>&database=dev&area=<area>

  1. Add the connection in Secrets and techniques Supervisor utilizing the next command within the AWS Command Line Interface (AWS CLI).

This will also be finished from the Secrets and techniques Supervisor console. This can be added in Secrets and techniques Supervisor as plaintext.

aws secretsmanager create-secret --name airflow/connections/secrets_redshift_connection --description "Apache Airflow to Redshift Cluster" --secret-string "redshift://admin:<password>@sample-workgroup.<account_id>.us-east-1.redshift-serverless.amazonaws.com:5439?role_arn=<MWAA position ARN>&database=dev&area=us-east-1" --region=us-east-1

Use the connection airflow/connections/secrets_redshift_connection within the DAG. When the DAG is run, it should search for this connection and retrieve the secrets and techniques from Secrets and techniques Supervisor. In case of RedshiftDataOperator, go the secret_arn as a parameter as an alternative of connection identify.

It’s also possible to add secrets and techniques utilizing the Secrets and techniques Supervisor console as key-value pairs.

  1. Add one other secret in Secrets and techniques Supervisor in and reserve it as airflow/connections/redshift_conn_test.

Create an Airflow connection via the metadata database

It’s also possible to create connections within the UI. On this case, the connection particulars can be saved in an Airflow metadata database. If the Amazon MWAA atmosphere will not be configured to make use of the Secrets and techniques Supervisor backend, it should examine the metadata database for the worth and return that. You possibly can create an Airflow connection utilizing the UI, AWS CLI, or API. On this part, we present the way to create a connection utilizing the Airflow UI.

  1. For Connection Id, enter a reputation for the connection.
  2. For Connection Sort, select Amazon Redshift.
  3. For Host, enter the Redshift endpoint (with out port and database) for Redshift Serverless.
  4. For Database, enter dev.
  5. For Consumer, enter your admin person identify.
  6. For Password, enter your password.
  7. For Port, use port 5439.
  8. For Further, set the area and timeout parameters.
  9. Check the connection, then save your settings.

Create and run a DAG

On this part, we describe the way to create a DAG utilizing varied parts. After you create and run the DAG, you’ll be able to confirm the outcomes by querying Redshift tables and checking the goal S3 buckets.

Create a DAG

In Airflow, information pipelines are outlined in Python code as DAGs. We create a DAG that consists of varied operators, sensors, connections, duties, and guidelines:

  • The DAG begins with searching for supply information within the S3 bucket sample-inp-bucket-etl-<username> below Account A for the present day utilizing S3KeySensor. S3KeySensor is used to attend for one or a number of keys to be current in an S3 bucket.
    • For instance, our S3 bucket is partitioned as s3://bucket/merchandise/YYYY/MM/DD/, so our sensor ought to examine for folders with the present date. We derived the present date within the DAG and handed this to S3KeySensor, which seems to be for any new information within the present day folder.
    • We additionally set wildcard_match as True, which allows searches on bucket_key to be interpreted as a Unix wildcard sample. Set the mode to reschedule in order that the sensor activity frees the employee slot when the factors will not be met and it’s rescheduled at a later time. As a finest observe, use this mode when poke_interval is greater than 1 minute to stop an excessive amount of load on a scheduler.
  • After the file is on the market within the S3 bucket, the AWS Glue crawler runs utilizing GlueCrawlerOperator to crawl the S3 supply bucket sample-inp-bucket-etl-<username> below Account A and updates the desk metadata below the products_db database within the Information Catalog. The crawler makes use of the AWS Glue position and Information Catalog database that had been created within the earlier steps.
  • The DAG makes use of GlueCrawlerSensor to attend for the crawler to finish.
  • When the crawler job is full, GlueJobOperator is used to run the AWS Glue job. The AWS Glue script identify (together with location) and is handed to the operator together with the AWS Glue IAM position. Different parameters like GlueVersion, NumberofWorkers, and WorkerType are handed utilizing the create_job_kwargs parameter.
  • The DAG makes use of GlueJobSensor to attend for the AWS Glue job to finish. When it’s full, the Redshift staging desk merchandise can be loaded with information from the S3 file.
  • You possibly can connect with Amazon Redshift from Airflow utilizing three totally different operators:
    • PythonOperator.
    • SQLExecuteQueryOperator, which makes use of a PostgreSQL connection and redshift_default because the default connection.
    • RedshiftDataOperator, which makes use of the Redshift Information API and aws_default because the default connection.

In our DAG, we use SQLExecuteQueryOperator and RedshiftDataOperator to indicate the way to use these operators. The Redshift saved procedures are run RedshiftDataOperator. The DAG additionally runs SQL instructions in Amazon Redshift to delete the info from the staging desk utilizing SQLExecuteQueryOperator.

As a result of we configured our Amazon MWAA atmosphere to search for connections in Secrets and techniques Supervisor, when the DAG runs, it retrieves the Redshift connection particulars like person identify, password, host, port, and Area from Secrets and techniques Supervisor. If the connection will not be present in Secrets and techniques Supervisor, the values are retrieved from the default connections.

In SQLExecuteQueryOperator, we go the connection identify that we created in Secrets and techniques Supervisor. It seems to be for airflow/connections/secrets_redshift_connection and retrieves the secrets and techniques from Secrets and techniques Supervisor. If Secrets and techniques Supervisor will not be arrange, the connection created manually (for instance, redshift-conn-id) will be handed.

In RedshiftDataOperator, we go the secret_arn of the airflow/connections/redshift_conn_test connection created in Secrets and techniques Supervisor as a parameter.

  • As closing activity, RedshiftToS3Operator is used to unload information from the Redshift desk to an S3 bucket sample-opt-bucket-etl in Account B. airflow/connections/redshift_conn_test from Secrets and techniques Supervisor is used for unloading the info.
  • TriggerRule is about to ALL_DONE, which allows the following step to run in spite of everything upstream duties are full.
  • The dependency of duties is outlined utilizing the chain() operate, which permits for parallel runs of duties if wanted. In our case, we wish all duties to run in sequence.

The next is the entire DAG code. The dag_id ought to match the DAG script identify, in any other case it gained’t be synced into the Airflow UI.

from datetime import datetime
from airflow import DAG 
from airflow.decorators import activity
from airflow.fashions.baseoperator import chain
from airflow.suppliers.amazon.aws.sensors.s3 import S3KeySensor
from airflow.suppliers.amazon.aws.operators.glue import GlueJobOperator
from airflow.suppliers.amazon.aws.operators.glue_crawler import GlueCrawlerOperator
from airflow.suppliers.amazon.aws.sensors.glue import GlueJobSensor
from airflow.suppliers.amazon.aws.sensors.glue_crawler import GlueCrawlerSensor
from airflow.suppliers.amazon.aws.operators.redshift_data import RedshiftDataOperator
from airflow.suppliers.widespread.sql.operators.sql import SQLExecuteQueryOperator
from airflow.suppliers.amazon.aws.transfers.redshift_to_s3 import RedshiftToS3Operator
from airflow.utils.trigger_rule import TriggerRule


dag_id = "data_pipeline"
vYear = datetime.right now().strftime("%Y")
vMonth = datetime.right now().strftime("%m")
vDay = datetime.right now().strftime("%d")
src_bucket_name = "sample-inp-bucket-etl-<username>"
tgt_bucket_name = "sample-opt-bucket-etl-<username>"
s3_folder="merchandise"
#Please change the variable with the glue_role_arn
glue_role_arn_key = "arn:aws:iam::<account_id>:position/<Glue-role>"
glue_crawler_name = "merchandise"
glue_db_name = "products_db"
glue_job_name = "sample_glue_job"
glue_script_location="s3://aws-glue-assets-<account_id>-<area>/scripts/sample_glue_job.py"
workgroup_name = "sample-workgroup"
redshift_table = "products_f"
redshift_conn_id_name="secrets_redshift_connection"
db_name = "dev"
secret_arn="arn:aws:secretsmanager:us-east-1:<account_id>:secret:airflow/connections/redshift_conn_test-xxxx"
poll_interval = 10

@activity
def get_role_name(arn: str) -> str:
    return arn.cut up("/")[-1]

@activity
def get_s3_loc(s3_folder: str) -> str:
    s3_loc  = s3_folder + "/yr=" + vYear + "/month=" + vMonth + "/day=" + vDay + "/*.csv"
    return s3_loc

with DAG(
    dag_id=dag_id,
    schedule="@as soon as",
    start_date=datetime(2021, 1, 1),
    tags=["example"],
    catchup=False,
) as dag:
    role_arn = glue_role_arn_key
    glue_role_name = get_role_name(role_arn)
    s3_loc = get_s3_loc(s3_folder)


    # Verify for brand spanking new incremental information in S3 supply/enter bucket
    sensor_key = S3KeySensor(
        task_id="sensor_key",
        bucket_key=s3_loc,
        bucket_name=src_bucket_name,
        wildcard_match=True,
        #timeout=18*60*60,
        #poke_interval=120,
        timeout=60,
        poke_interval=30,
        mode="reschedule"
    )

    # Run Glue crawler
    glue_crawler_config = {
        "Identify": glue_crawler_name,
        "Function": role_arn,
        "DatabaseName": glue_db_name,
    }

    crawl_s3 = GlueCrawlerOperator(
        task_id="crawl_s3",
        config=glue_crawler_config,
    )

    # GlueCrawlerOperator waits by default, setting as False to check the Sensor beneath.
    crawl_s3.wait_for_completion = False

    # Look ahead to Glue crawler to finish
    wait_for_crawl = GlueCrawlerSensor(
        task_id="wait_for_crawl",
        crawler_name=glue_crawler_name,
    )

    # Run Glue Job
    submit_glue_job = GlueJobOperator(
        task_id="submit_glue_job",
        job_name=glue_job_name,
        script_location=glue_script_location,
        iam_role_name=glue_role_name,
        create_job_kwargs={"GlueVersion": "4.0", "NumberOfWorkers": 10, "WorkerType": "G.1X"},
    )

    # GlueJobOperator waits by default, setting as False to check the Sensor beneath.
    submit_glue_job.wait_for_completion = False

    # Look ahead to Glue Job to finish
    wait_for_job = GlueJobSensor(
        task_id="wait_for_job",
        job_name=glue_job_name,
        # Job ID extracted from earlier Glue Job Operator activity
        run_id=submit_glue_job.output,
        verbose=True,  # prints glue job logs in airflow logs
    )

    wait_for_job.poke_interval = 5

    # Execute the Saved Process in Redshift Serverless utilizing Information Operator
    execute_redshift_stored_proc = RedshiftDataOperator(
        task_id="execute_redshift_stored_proc",
        database=db_name,
        workgroup_name=workgroup_name,
        secret_arn=secret_arn,
        sql="""CALL sp_products();""",
        poll_interval=poll_interval,
        wait_for_completion=True,
    )

    # Execute the Saved Process in Redshift Serverless utilizing SQL Operator
    delete_from_table = SQLExecuteQueryOperator(
        task_id="delete_from_table",
        conn_id=redshift_conn_id_name,
        sql="DELETE FROM merchandise;",
        trigger_rule=TriggerRule.ALL_DONE,
    )

    # Unload the info from Redshift desk to S3
    transfer_redshift_to_s3 = RedshiftToS3Operator(
        task_id="transfer_redshift_to_s3",
        s3_bucket=tgt_bucket_name,
        s3_key=s3_loc,
        schema="PUBLIC",
        desk=redshift_table,
        redshift_conn_id=redshift_conn_id_name,
    )

    transfer_redshift_to_s3.trigger_rule = TriggerRule.ALL_DONE

    #Chain the duties to be executed
    chain(
        sensor_key,
        crawl_s3,
        wait_for_crawl,
        submit_glue_job,
        wait_for_job,
        execute_redshift_stored_proc,
        delete_from_table,
        transfer_redshift_to_s3
        )
    

Confirm the DAG run

After you create the DAG file (change the variables within the DAG script) and add it to the s3://sample-airflow-instance/dags folder, will probably be routinely synced with the Airflow UI. All DAGs seem on the DAGs tab. Toggle the ON choice to make the DAG runnable. As a result of our DAG is about to schedule="@as soon as", that you must manually run the job by selecting the run icon below Actions. When the DAG is full, the standing is up to date in inexperienced, as proven within the following screenshot.

Within the Hyperlinks part, there are alternatives to view the code, graph, grid, log, and extra. Select Graph to visualise the DAG in a graph format. As proven within the following screenshot, every coloration of the node denotes a particular operator, and the colour of the node define denotes a particular standing.

Confirm the outcomes

On the Amazon Redshift console, navigate to the Question Editor v2 and choose the info within the products_f desk. The desk needs to be loaded and have the identical variety of data as S3 information.

On the Amazon S3 console, navigate to the S3 bucket s3://sample-opt-bucket-etl in Account B. The product_f information needs to be created below the folder construction s3://sample-opt-bucket-etl/merchandise/YYYY/MM/DD/.

Clear up

Clear up the sources created as a part of this publish to keep away from incurring ongoing expenses:

  1. Delete the CloudFormation stacks and S3 bucket that you simply created as conditions.
  2. Delete the VPCs and VPC peering connections, cross-account insurance policies and roles, and secrets and techniques in Secrets and techniques Supervisor.

Conclusion

With Amazon MWAA, you’ll be able to construct complicated workflows utilizing Airflow and Python with out managing clusters, nodes, or every other operational overhead sometimes related to deploying and scaling Airflow in manufacturing. On this publish, we confirmed how Amazon MWAA offers an automatic option to ingest, rework, analyze, and distribute information between totally different accounts and companies inside AWS. For extra examples of different AWS operators, check with the next GitHub repository; we encourage you to be taught extra by making an attempt out a few of these examples.


Concerning the Authors


Radhika Jakkula is a Large Information Prototyping Options Architect at AWS. She helps clients construct prototypes utilizing AWS analytics companies and purpose-built databases. She is a specialist in assessing big selection of necessities and making use of related AWS companies, large information instruments, and frameworks to create a sturdy structure.

Sidhanth Muralidhar is a Principal Technical Account Supervisor at AWS. He works with giant enterprise clients who run their workloads on AWS. He’s captivated with working with clients and serving to them architect workloads for prices, reliability, efficiency, and operational excellence at scale of their cloud journey. He has a eager curiosity in information analytics as nicely.

Related Articles

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Latest Articles