Skip to content

Redshift へのデータ転送

Amazon Redshift は、Amazon S3 と並んでAmazon Web Services 上で動作する、評判の良いデータウェアハウスです。Currents の Braze データは Redshift に直接転送しやすい構造になっています。

ETL プロセス経由で Amazon S3 から Redshift にデータを転送する方法の詳細については、Currents の例がある GitHub リポジトリを参照してください。

Sample Currents S3 to Redshift Loader

Overview

The basic premise of the code within s3loader.py involves using a separate manifest table in the same Redshift database to keep track of the files that have already been copied. The general structure is as follows:

  1. List all of the files in S3, identify the new files since the last time you’ve ran s3loader.py by comparing the list with the contents in the manifest table.
  2. Create a manifest file containing the new files.
  3. Execute a COPY query to copy the new files from S3 to Redshift using the manifest file.
  4. Insert the names of the files that are copied into the separate manifest table in Redshift.
  5. Commit.

Dependencies

You must install the AWS Python SDK and Psycopg in order to run the Loader:

1
2

pip install boto3
pip install psycopg2

Permissions

Redshift Role with S3 Read Access

If you have not done so, follow the AWS Documentation to create a Role that is able to execute COPY commands on your files in S3: http://docs.aws.amazon.com/redshift/latest/gsg/rs-gsg-create-an-iam-role.html.

Redshift VPC Inbound Rules

If your Redshift cluster is in a VPC, you must configure the VPC to allow connections from the server that you are running the S3 Loader. Go into your Redshift Cluster and select the VPC Security Groups entry that you want the Loader to connect into. Next, add a new Inbound Rule: Type = Redshift, Protocol = TCP, Port = the port for your cluster, Source = Anywhere (or wherever the Loader is running)/

IAM User with S3 Full Access

The S3 Loader requires read access to the files containing your Currents data, and full access to the location for the manifest files that it generates for the Redshift COPY commands. The easiest thing to do to accomplish this is to create a new IAM User with the AmazonS3FullAccess permissions: https://console.aws.amazon.com/iam/home#/users. Be sure to save the credentials as you will need to pass them to the Loader.

You can pass the access credentials to the Loader via environment variables, the shared credential file (~/.aws/credentials), or the AWS Config File (http://boto3.readthedocs.io/en/latest/guide/configuration.html#configuring-credentials). Alternatively you may include them directly in the Loader by assigning them to the aws_access_key_id and the aws_secret_access_key fields within an S3LoadJob object, but we do not recommend that you hard code the credentials within your source code.

Usage

Sample Usage

Below is a sample program that loads data for the users.behaviors.app.NewsFeedImpression event from S3 to the tutorial_newsfeed_impression table in Redshift.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47

if __name__ == '__main__':
  host = '<your_host>.redshift.amazonaws.com'
  port = 5439
  database = '<your_db>'
  user = '<your_db_user>'
  password = '<your_db_password>'
  role = '<your_redshift_role_with_s3_read_access>'

  # Do not hard code these credentials.
  aws_access_key_id = None
  aws_secret_access_key = None

  '''
  News Feed Impression Avro fields:
  [
    {"name": "id", "type": "string"}, 
    {"name": "user_id", "type": "string"},
    {"name": "external_user_id", "type": ["null", "string"], "default": null},
    {"name": "app_id", "type": "string"}, 
    {"name": "time", "type": "int"},
    {"name": "platform", "type": ["null", "string"], "default": null},
    {"name": "os_version", "type": ["null", "string"], "default": null},
    {"name": "device_model", "type": ["null", "string"], "default": null}
  ]  
  '''
  print 'Loading News Feed Impression...'
  newsfeed_impression_s3_bucket = '<your_currents_bucket>'
  newsfeed_impression_s3_prefix = '<your_prefix_to_event_type=users.behaviors.app.NewsFeedImpression>'
  newsfeed_impression_redshift_table = 'tutorial_newsfeed_impression'
  newsfeed_impression_redshift_column_def = [
    ('id', 'text'),
    ('user_id', 'text'),
    ('external_user_id', 'text'), 
    ('app_id', 'text'),
    ('time', 'integer'),
    ('platform', 'text'),
    ('os_version', 'text'),
    ('device_model', 'text')
  ]

  newsfeed_impression_redshift = RedshiftEndpoint(host, port, database, user, password, 
    newsfeed_impression_redshift_table, newsfeed_impression_redshift_column_def)
  newsfeed_impression_s3 = S3Endpoint(newsfeed_impression_s3_bucket, newsfeed_impression_s3_prefix)

  newsfeed_impression_job = S3LoadJob(newsfeed_impression_redshift, newsfeed_impression_s3, role, 
    aws_access_key_id=aws_access_key_id, aws_secret_access_key=aws_secret_access_key)
  newsfeed_impression_job.perform()

Credentials

In order to run the Loader, you must first provide the host, port, and database of your Redshift cluster as well as the user and password of a Redshift user that can run COPY queries. Additionally, you must provide the ARN of the Redshift Role with S3 Read Access that you’ve created from a previous section.

1
2
3
4
5
6

  host = '<your_host>.redshift.amazonaws.com'
  port = 5439
  database = '<your_db>'
  user = '<your_db_user>'
  password = '<your_db_password>'
  role = '<your_redshift_role_with_s3_read_access>'

Job Configuration

You must provide the S3 Bucket and Prefix of your event files, as well as the Redshift table name that you want to COPY into.

In addition, in order to COPY Avro files with the “auto” option as required by the Loader, the column definition in your Redshift table must match the field names in the Avro schema as shown in the Sample program, with the appropriate type mapping (e.g. “string” to “text”, “int” to “integer”).

You may also pass in a batch_size option to the Loader if you deem that it takes too long to copy all of the files at once. Passing a batch_size allows the Loader to incrementally copy and commit one batch at a time without having to copy everything at the same time. The time that it takes to load one batch depends on the batch_size as well as the size of your files and the size of your Redshift cluster.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

  '''
  News Feed Impression Avro fields:
  [
    {"name": "id", "type": "string"}, 
    {"name": "user_id", "type": "string"},
    {"name": "external_user_id", "type": ["null", "string"], "default": null},
    {"name": "app_id", "type": "string"}, 
    {"name": "time", "type": "int"},
    {"name": "platform", "type": ["null", "string"], "default": null},
    {"name": "os_version", "type": ["null", "string"], "default": null},
    {"name": "device_model", "type": ["null", "string"], "default": null}
  ]  
  '''
  newsfeed_impression_s3_bucket = '<your_currents_bucket>'
  newsfeed_impression_s3_prefix = '<your_prefix_to_event_type=users.behaviors.app.NewsFeedImpression>'
  newsfeed_impression_redshift_table = 'tutorial_newsfeed_impression'
  newsfeed_impression_redshift_column_def = [
    ('id', 'text'),
    ('user_id', 'text'),
    ('external_user_id', 'text'), 
    ('app_id', 'text'),
    ('time', 'integer'),
    ('platform', 'text'),
    ('os_version', 'text'),
    ('device_model', 'text')
  ]
  newsfeed_impression_batch_size = 1000
Github   GitHub でこのページを編集
「このページはどの程度役に立ちましたか?」
New Stuff!