Skip to content

Greenmask 0.2.0

This is one of the biggest releases since Greenmask was founded. We've been in close contact with our users, gathering feedback, and working hard to make Greenmask more flexible, reliable, and user-friendly.

This major release introduces exciting new features such as database subsetting, pgzip support, restoration in topological order, and refactored transformers, significantly enhancing Greenmask's flexibility to better meet business needs. It also includes several fixes and improvements.

Preface

This release is a major milestone that significantly expands Greenmask's functionality, transforming it into a simple, extensible, and reliable solution for database security, data anonymization, and everyday operations. Our goal is to create a core system that can serve as a foundation for comprehensive dynamic staging environments and robust data security.

Notable changes

  • PostgreSQL 17 support - revised ported library to support PostgreSQL 17

  • Database Subset - a new feature that allows you to define a subset of the database, allowing you to scale down the dump size (#110). This is robust for multipurpose and especially useful for testing and development environments. It supports:

    • References with NULL values - generate the LEFT JOIN query for the FK reference with NULL values to include them in the subset.
    • Supports virtual references (virtual foreign keys) - create a logical FK in Greenmask that will be used for subset dependencies graph. The virtual reference can be defined for a column or an expression, allowing you to get the value from JSON and similar.
    • Supports circular references - Greenmask will automatically resolve circular dependencies in the subset by generating a recursive query. The query is generated with integrity checks of the subset ensuring that the data gathered from circular dependencies is consistent.
    • Fully covered with documentation including troubleshooting and examples.
    • Supports FK and PK that have more than one column (or expression).
    • Multi-cycles resolution in one strong connected component (SCC) is supported - Greenmask will generate a recursive query for the SCC whether it is a single cycle or multiple cycles, making the subset system universal for any database schema.
    • Supports polymorphic relationships - You can define a virtual reference for a table with polymorphic references using polymorphic_exprs attribute and use greenmask to generate a subset for such tables.
  • pgzip support for faster compression and decompression — setting --pgzip can speed up the dump and restoration processes through parallel compression. In some tests, it shows up to 5x faster dump and restore operations.

  • Restoration in topological order - This flag ensures that dependent tables are not restored until the tables they depend on have been restored. This is useful when you want to be notified of errors as immediately as possible without waiting for the entire table to be restored.
  • Insert format restoration - For a flexible restoration process, Greenmask now supports data restoration in the INSERT format. It generates the insert statements based on COPY records from the dump. You do not need to re-dump your data to use this feature; it can be defined in the restore command. The list of new features related to the INSERT format:

    • Generate INSERT statements with the **ON CONFLICT DO NOTHING** clause if the flag --on-conflict-do-nothing is set.
    • Error exclusion list in the config to skip certain errors and continue inserting subsequent rows from the dump.
    • Use cases - incremental dump and restoration for logical data. For example, if you have a database, and you want to insert data periodically from another source, this can be used together with the database subset and transformations to catch up the target database.
  • Restore data batching (#173) - By default, the COPY protocol returns the error only on transaction commit. To override this behavior, use the --batch-size flag to specify the number of rows to insert in a single batch during the COPY command. This is useful when you want to control the transaction size and commit.

  • Introduced keep_null parameter for RandomPerson transformer.

  • Introduced dynamic parameters in the transformers

    • Most transformers now support dynamic parameters where applicable.
    • Dynamic parameters are strictly enforced. If you need to cast values to another type, Greenmask provides templates and predefined cast functions accessible via cast_to. These functions cover frequent operations such as UnixTimestampToDate and IntToBool.
  • The transformation logic has been significantly refactored, making transformers more customizable and flexible than before.
  • Introduced transformation engines

    • random - generates transformer values based on pseudo-random algorithms.
    • hash - generates transformer values using hash functions. Currently, it utilizes sha3 hash functions, which are secure but perform slowly. In the stable release, there will be an option to choose between sha3 and SipHash.
  • Introduced static parameters value template

  • Dumps retention management - Introduced retention parameters (#201) for the delete command. Introduced two new statuses: failed and in progress. A dump is considered failed if it lacks a "done" heartbeat or if the last heartbeat timestamp exceeds 30 minutes. The delete command now supports the following retention parameters:

    • --dry-run: Runs the deletion operation in test mode with verbose output, without actually deleting anything.
    • --before-date 2024-08-27T23:50:54+00:00: Deletes dumps older than the specified date. The date must be provided in RFC3339Nano format, for example: 2021-01-01T00:00:00Z.
    • --retain-recent 10: Retains the N most recent dumps, where N is specified by the user.
    • --retain-for 1w2d3h4m5s6ms7us8ns: Retains dumps for the specified duration. The format supports weeks (w), days (d), hours (h), minutes (m), seconds (s), milliseconds (ms), microseconds (us), and nanoseconds (ns).
    • --prune-failed: Prunes (removes) all dumps that have failed.
    • --prune-unsafe: Prunes dumps with "unknown-or-failed" statuses. This option only works in conjunction with --prune-failed.
  • Docker image mirroring into the GitHub Container Registry

Core

  • Introduced the Parametrizer interface, now implemented for both dynamic and static parameters.
  • Renamed most of the toolkit types for enhanced clarity and comprehensive documentation coverage.
  • Refactored the Driver initialization logic.
  • Added validation warnings for overridden types in the Driver.
  • Migrated existing built-in transformers to utilize the new Parametrizer interface.
  • Implemented a new abstraction, TransformationContext, as the first step towards enabling new feature transformation conditions (#34).
  • Optimized most transformers for performance in both dynamic and static modes. While dynamic mode offers flexibility, static mode ensures performance remains high. Using only the necessary transformation features helps keep transformation time predictable.

Transformers

  • RandomEmail - Introduces a new transformer that supports both random and deterministic engines. It allows for flexible email value generation; you can use column values in the template and choose to keep the original domain or select any from the domains parameter.

  • NoiseDate, NoiseFloat, NoiseInt - These transformers support both random and deterministic engines, offering dynamic mode parameters that control the noise thresholds within the min and max range. Unlike previous implementations which used a single ratio parameter, the new release features min_ratio and max_ratio parameters to define noise values more precisely. Utilizing the hash engine in these transformers enhances security by complicating statistical analysis for attackers, especially when the same salt is used consistently over long periods.

  • NoiseNumeric - A newly implemented transformer, sharing features with NoiseInt and NoiseFloat, but specifically designed for numeric values (large integers or floats). It provides a decimal parameter to handle values with fractions.

  • RandomChoice - Now supports the hash engine

  • RandomDate, RandomFloat, RandomInt - Now enhanced with hash engine support. Threshold parameters min and max have been updated to support dynamic mode, allowing for more flexible configurations.

  • RandomNumeric - A new transformer specifically designed for numeric types (large integers or floats), sharing similar features with RandomInt and RandomFloat, but tailored for handling huge numeric values.

  • RandomString - Now supports hash engine mode

  • RandomUnixTimestamp - This new transformer generates Unix timestamps with selectable units (second, millisecond, microsecond, nanosecond). Similar in function to RandomDate, it supports the hash engine and dynamic parameters for min and max thresholds, with the ability to override these units using min_unit and max_unit parameters.

  • RandomUuid - Added hash engine support

  • RandomPerson - Implemented a new transformer that replaces RandomName, RandomLastName, RandomFirstName, RandomFirstNameMale, RandomFirstNameFemale, RandomTitleMale, and RandomTitleFemale. This new transformer offers enhanced customizability while providing similar functionalities as the previous versions. It generates personal data such as FirstName, LastName, and Title, based on the provided gender parameter, which now supports dynamic mode. Future minor versions will allow for overriding the default names database.

  • Added tsModify - a new template function for time.Time objects modification

  • Introduced a new RandomIp transformer capable of generating a random IP address based on the specified netmask.

  • Added a new RandomMac transformer for generating random Mac addresses.

  • Deleted transformers include RandomMacAddress, RandomIPv4, RandomIPv6, RandomUnixTime, RandomTitleMale, RandomTitleFemale, RandomFirstName, RandomFirstNameMale, RandomFirstNameFemale, RandomLastName, and RandomName due to the introduction of more flexible and unified options.

Fixes and improvements

  • Fixed validate command with the --table flag, which had the wrong order of the table name representation {{ table_name }}.{{ schema }} instead of {{ schema }}.{{ table_name }}.
  • Fixed Row.SetColumn out of range validation.
  • Fixed restoreWorker panic caused when the worker received an error from pgx.
  • Fixed error handling in the restore command.
  • Fixed restore jobs now start a transaction for each table restoration and commit it after the table restoration is done.
  • Fixed --exit-on-error works incorrectly in the restore command. Now, the --exit-on-error flag works correctly with the data section.
  • Fixed transaction rollback in the validate command.
  • Fixed typo in documentation.
  • Fixed a CI/CD bug related to retrieving current tags.
  • Fixed the Docker image tag for latest to exclude specific keywords.
  • Fixed a case where the hashing value was not set for each column in the RandomPerson transformer.
  • Fixed original email value parsing conditions.
  • Subset docs revision.
  • Fixes a case where data entries were excluded by exclusion parameters such as --exclude-table, --table, etc.
  • Fixed zero bytes that were written in the buffer due to the wrong buffer limit in the Email transformer.
  • Fixed a case where the overridden type of column via columns_type_override did not work.
  • Fixed a case where an unknown option provided in the config was just ignored instead of throwing an error.
  • Fixed a case where min and max parameter values were ignored in transformers NoiseDate, NoiseNumeric, NoiseFloat, NoiseInt, RandomNumeric, RandomFloat, and RandomInt.
  • Fixed TOC entry COPY restoration statement - added missing newline and semicolon. Now backward pg_dump call pg_restore 1724504511561 --file 1724504511561.sql is backward compatible and works as expected.
  • Fixed a case where dump/restore fails when masking tables with a generated column.
  • Updated go version (v1.22) and dependencies
  • Revised installation section of doc
  • PostgreSQL 17 support - revised ported library to support PostgreSQL 17
  • Fixed integration tests - reset the go test cache on each iteration
  • Push docker images to ghcr.io registry
  • A bunch of refactoring and code cleanup to make the codebase more maintainable and readable.

Full Changelog: v0.1.14...v0.2.0

Feel free to reach out to us if you have any questions or need assistance: