Getting started with Amazon Glue Data Quality for the Data Catalog - Amazon Glue
PrerequisitesStep-by-step exampleGenerating rule recommendationsMonitoring rule recommendationsEditing recommended rulesetsCreating a new rulesetRunning a ruleset to evaluate data qualityViewing the data quality score and resultsRelated topics
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Getting started with Amazon Glue Data Quality for the Data Catalog

This getting started section provides instructions to help you get started with Amazon Glue Data Quality on the Amazon Glue console. You'll learn how to complete essential tasks such as generating data quality rule recommendations and evaluating a ruleset against your data.

Prerequisites

Before you use Amazon Glue Data Quality, you should be familiar with using the Data Catalog and crawlers in Amazon Glue. With Amazon Glue Data Quality, you can evaluate quality for tables in a Data Catalog database. You also need the following:

  • A table in the Data Catalog to evaluate your data quality ruleset against.

  • An IAM role for Amazon Glue that you supply when you generate rule recommendations or run a data quality task. This role must have permission to access resources that various Amazon Glue Data Quality processes require to run on your behalf. These resources include Amazon Glue, Amazon S3, and CloudWatch. To view example policies that include the minimum permissions for Amazon Glue Data Quality, see Example IAM policies.

    To learn more about IAM roles for Amazon Glue, see Create an IAM policy for the Amazon Glue service and Create an IAM role for the Amazon Glue service. You can also view a list of all Amazon Glue permissions that are specific to data quality in Authorization for Amazon Glue Data Quality actions.

  • A database with at least one table that contains a variety of data. The table used in this tutorial is named yyz-tickets, with the table tickets. This data is a collection of publicly available information from the City of Toronto for parking citations. If you create your own table, make sure that it's populated with a variety of valid data to get the best set of recommended rules.

Step-by-step example

For a step-by-step example with sample datasets, see the Amazon Glue Data Quality blog post.

Generating rule recommendations

Rule recommendations make it easy to get started with data quality without writing code. With Amazon Glue Data Quality, you can analyze your data, identify rules, and create a ruleset that you can evaluate in a data quality task. Recommendation runs are automatically deleted after 90 days.

To generate data quality rule recommendations

  1. Open the Amazon Glue console at https://console.amazonaws.cn/glue/.

  2. Choose Tables in the navigation pane. Then choose the table that you want to generate data quality rule recommendations for.

  3. On the table details page, select the Data quality tab to access Amazon Glue Data Quality rules and settings for your table.

  4. On the Data quality tab, choose Add rules and monitor data quality.

  5. On the Ruleset builder page, an alert at the top of the page will prompt you to start a recommendation task if there are no rule recommendation runs.

  6. Choose Recommend rules to open the modal and input your parameters for the recommendation task.

  7. Choose an IAM role with access to Amazon Glue. This role must have permission to access resources that various Amazon Glue Data Quality processes require to run on your behalf.

  8. After the fields are completed according to your preferences, choose Recommend rules to start the recommendation task run. If recommendation runs are in progress or completed, you can manage your runs in this alert. You might need to refresh the alert to view the status change. Completed and in-progress recommendation task runs appear in the Run history page that lists all recommendation runs for the past 90 days.

What the recommended rules mean

Amazon Glue Data Quality generates rules based on the data from each column of the input table. It uses the rules to identify potential boundaries where data can be filtered to maintain quality requirements. The following list of generated rules includes examples that are useful for understanding what the rules mean and what they might do when applied to your data.

For a full list of the generated Data Quality Definition Language (DQDL) rule types, see DQDL rule type reference.

  • IsComplete "SET_FINE_AMOUNT" –The IsComplete rule verifies that the column is filled in for any given row. Use this rule to tag columns as non-optional in data.

  • Uniqueness "TICKET_NUMBER" > 0.95 – The Uniqueness rule verifies that the data within the column meets some threshold of uniqueness. In this example, the data that populates any given row for "TICKET_NUMBER" was determined to be at most 95% identical in content to all other rows, which suggests this rule.

  • ColumnValues "PROVINCE" in ["ON", "QC", "AB", "NY",...] – The ColumnValues rule defines valid values for the column, based on existing column contents. In this example, the data for each row is a 2-letter license code plate for a state or province.

  • ColumnLength "INFRACTION_DESCRIPTION" between 15 and 31 – The ColumnLength rule enforces a length restriction on a column's data. This rule is generated from the sample data based on the minimum and maximum recorded lengths for a column of strings.

Monitoring rule recommendations

When data quality rule recommendations are running, the Add rules and monitor data quality page displays information and additional actions that you can take in the top bar.

When rule recommendations are in progress, you can choose Stop run before the recommendation task is complete. While the task is in progress, you will see the status, in progress, and the date and time when the run started.

When the rule recommendations are complete, the rule recommendation bar displays the number of rules recommended, the status of the last recommendation run, and the date and timestamp when it finished.

You can add the recommended rules by choosing Insert Rule Recommendation. To view previously recommended rules, select a specific date. To run a new recommendation, choose More actions, and then choose Recommended rules.

Set default settings by choosing Manage user settings. You can set the default path for Amazon S3 to store rulesets or to set up a default role to run the Data Catalog.

Editing recommended rulesets

Because Amazon Glue Data Quality generates rules based on existing data that you have available, you might see some unexpected or undesirable rules in the automated suggestions. In order to get the most out of the recommended rulesets, you need to evaluate and modify them. For this step of the tutorial, you take the rules generated in the previous step and adjust them to enforce more restrictive qualities on some data. You also relax other rules to ensure that correct, unique data can be added later.

Edit a suggested ruleset

  1. In the Amazon Glue console, choose Data Catalog, and then choose Databases tables in the navigation pane. Choose the table tickets.

  2. On the table details page, choose the Data quality tab to access Amazon Glue Data Quality rules and settings for the table.

  3. In the Rulesets section, select the ruleset generated in Generating rule recommendations.

  4. Choose Actions, and then choose Edit in the console window. The ruleset editor loads in the console. It includes an editing pane for your rules and a quick reference for DQDL.

  5. Remove line 2 of the script. This relaxes the requirement that the database size is constrained within a certain number of rows. After the edit, your file should contain the following on lines 1–3:

    Rules = [
    IsComplete "TAG_NUMBER_MASKED",
    ColumnLength "TAG_NUMBER_MASKED" between 6 and 9,

  6. Remove line 25 of the script. This relaxes the requirement that 96% of recorded provinces are ON. After the edit, your file should contain the following from line 24 to the end of the ruleset:

    ColumnValues "PROVINCE" in ["ON", "QC", "AB", "NY", "AZ", "NS", "BC", "MI", "PQ", "MB", "PA", "FL", "SK", "NJ", "OH", "NB", "IL", "MA", "CA",
    "VA", "TX", "NF", "MD", "PE", "CT", "NC", "GA", "IN", "OR", "MN", "TN", "WI", "KY", "MO", "WA", "NH", "SC", "CO", "OK", "VT", "RI", "ME", "AL",
    "YT", "IA", "DE", "AR", "LA", "XX", "WV", "MT", "KS", "NT", "DC", "NV", "NE", "UT", "MS", "NM", "ID", "SD", "ND", "AK", "NU", "GO", "WY", "HI"],
ColumnLength "PROVINCE" = 2
]

  7. Change line 14 to the following:

    IsComplete "TIME_OF_INFRACTION",

    This strengthens the requirement on the column by limiting the database to only tickets that contain a recorded time of infraction. You should always consider tickets without a recorded time of infraction to be invalid data in this dataset. This is different than situations where partitioning or transformation might be more appropriate for further data use or inspection to determine a quality rule.

  8. Choose Update Rulesetat the bottom of the console page.

Creating a new ruleset

A ruleset is a group of data quality rules that you evaluate against your data. In the Amazon Glue console, you can author custom rulesets using Data Quality Definition Language (DQDL).

To create a data quality ruleset

  1. In the Amazon Glue console, choose Data Catalog, choose Databases, and then choose Tables in the navigation pane. Select the table tickets.

  2. Open the Data quality tab.

  3. In the Rulesets section, choose Create ruleset. The DQDL editor launches in the console. It has a text area for direct editing, and a quick reference for DQDL rules and the table schema.

  4. Start adding rules to the text area of the DQDL editor. You can either write rules directly from this tutorial, or use the DQDL rule builder feature of the data quality rules editor.

    Note
    How to use the DQDL rule builder

    1. Select a rule type from the list, and select the plus sign to insert example syntax into the editor pane.

    2. Exchange the placeholder column names with your own column names. Column names from the table are available in the Schema tab.

    3. Update the expression parameter as you see fit. For a full list of expressions that DQDL supports, see Expressions.

    As an example, the following rules are constraints for data validation of the ticket_number column in the tickets table. To add the following rules, use the DQDL rule builder or directly edit your ruleset: 

    IsComplete "ticket_number",
IsUnique "ticket_number",
ColumnValues "ticket_number" > 9000000000

  5. Provide a name for your new ruleset in the Ruleset name field.

  6. Choose Save ruleset.

Evaluating data quality across multiple datasets

You can set up data quality rules across multiple datasets using ReferentialIntegrity and DatasetMatch rulesets. ReferentialIntegrity checks to see if data in the primary dataset is present in other datasets.

To add a reference dataset, choose the Schema tab and then choose Update reference tables. You will be prompted to select a database and a table. You can add the table and then set up data quality rules. Rule types like AggregateMatch, RowCountMatch, ReferentialIntegrity, SchemaMatch, and DatasetMatch support the ability to perform data quality checks across multiple datasets.

Running a ruleset to evaluate data quality

When you run a data quality task, Amazon Glue Data Quality evaluates a ruleset against your data and calculates a data quality score. This score represents the percentage of data quality rules that passed for the input.

To run a data quality task

  1. In the Amazon Glue console, choose Data Catalog, choose Databases, and then choose Tables in the navigation pane. Select the table tickets.

  2. Choose the Data quality tab.

  3. In the Rulesets list, select the ruleset that you want to evaluate against the table. For this step, we recommend using a ruleset that you've written or modified already rather than generated rules. Choose Run.

  4. In the modal, choose your IAM role. This role must have permission to access resources that various Amazon Glue Data Quality processes require to run on your behalf. You can save the IAM role as the default or modify it by going to the Default Setting page.

  5. Under Data quality actions, choose whether you want to Publish metrics to Amazon CloudWatch. When this option is selected, Amazon Glue Data Quality publishes metrics that indicate the number of rules that passed and the number of rules that failed. To take action on metrics stored this way, you can use CloudWatch alarms. Key metrics are also published to Amazon EventBridge for you to set up alerts. For more information, see Setting up alerts, deployments, and scheduling.

  6. In Run Frequency, choose run on demand or schedule the ruleset. When you schedule a ruleset, you're prompted for a task name. The schedule will be created in Amazon EventBridge. You can edit your schedule in Amazon EventBridge.

  7. To save the data quality results in Amazon S3, select a Data quality results location. The IAM role that you previously selected for this task must have write access to this location.

  8. Under Additional Configurations, enter the Requested number of workers that you want Amazon Glue to allocate for your data quality task.

  9. You can optionally set up a filter at the data source. This helps you reduce the data that you're reading. You can also use a filter to run incremental validations by selecting partition information and passing them as parameters via API calls. To improve performance, you can provide a partition predicate.

  10. Choose Run. You should see your new task in the Data quality task runs list. When the Run status column for the task shows as Completed, you can view the quality score results. You might need to refresh your console window for the status to update correctly.

  11. To view the column for the data quality result details, choose the “+” icon to expand the ruleset. The results show you the rules that passed and failed in the evaluation, and what triggered the rule failure.

Viewing the data quality score and results

To see the latest run on all created rulesets

  1. In the Amazon Glue console, choose Tables in the navigation pane. Then choose the table that you want to run a data quality task for.

  2. Choose the Data quality tab.

  3. The Data quality snapshot shows a general trend of runs over time. The last 10 runs over all rulesets are displayed by default. To filter by ruleset, select the desired one from the dropdown list. If there are less than 10 runs, all the completed runs that are available are displayed.

  4. In the Data quality table, each ruleset with its latest run (if there is one) is shown, along with the score. Expanding the ruleset displays the rules that are in that ruleset, along with the rule results for that run.

To see the latest run on a particular ruleset

  1. In the Amazon Glue console, choose Tables in the navigation pane. Then choose the table that you want to run a data quality task for.

  2. Choose the Data quality tab.

  3. In the Data quality table, choose on a specific ruleset.

  4. On the Ruleset details page, choose the Run history tab.

    All of the evaluation runs for this particular ruleset are listed in the table within this tab. You can see the history of the scores and the status of the runs.

  5. To see more information about a particular run, choose the Run ID to go to the Evaluation run details page. On this page, you can see specifics about the run and more details about the status of individual rule results.