#amazon-s3 #s3 #postgresql #cdc #parquet #polars

dms-cdc-operator

The dms-cdc-operator is a Rust-based utility for comparing the state of a list of tables in an Amazon RDS database with data stored in Parquet files on Amazon S3, particularly useful for change data capture (CDC) scenarios

10 releases

0.1.14 Jun 11, 2024
0.1.13 Jun 11, 2024
0.1.9 May 29, 2024
0.1.0 Apr 26, 2024

#347 in Database interfaces

Download history 112/week @ 2024-04-21 295/week @ 2024-04-28 251/week @ 2024-05-05 181/week @ 2024-05-12 358/week @ 2024-05-19 376/week @ 2024-05-26 493/week @ 2024-06-02 370/week @ 2024-06-09 52/week @ 2024-06-16

1,315 downloads per month
Used in dms-cdc-operator-client

MIT license

87KB
2K SLoC

DMS-CDC-Operator

A Rust 🦀 tool for comparing Amazon RDS tables with Parquet files on Amazon S3, useful for change data capture (CDC)

Overview

The rust-dms-cdc-operator is a Rust-based tool that compares tables in an Amazon RDS (PostgreSQL) database with data migrated to Amazon S3 using AWS Database Migration Service (DMS). It's particularly useful for ensuring data consistency between the RDS database and Parquet files in S3, especially with change data capture (CDC) updates, since DMS validation with S3 as target isn't supported yet.

Features

  • Import a snapshot of the CDC parquet data stored in AWS S3 with date-based folder partitioning in a locally deployed Postgres
  • Specify a specific time range to replicate the S3 state on a Postgres DB
  • Restore the RDS state from S3 in case of data loss
  • Compare the state of a specific table in an Amazon RDS database with the data stored in Parquet files in the S3 bucket
  • Identify differences at the row level by modifying the validated chunk size
  • Use it as a library so as to integrate it in your projects, or as a client so as to use it as a standalone tool

Prerequisites

  • Your source DB is a PostgreSQL
  • You have a running AWS DMS task in FULL LOAD (or FULL LOAD + CDC) Mode
  • The target of the task is AWS S3 with:
    • Parquet formatted files
    • date-based folder partitioning
    • Additional column of Op injected by DMS

Installation (Client)

In order to use the tool as a client, you can use cargo.

The tool provides two features for running it, which are Inquire and Clap.

Using Clap

Usage: dms-cdc-operator-client validate [OPTIONS] --bucket-name <BUCKET_NAME> --s3-prefix <S3_PREFIX> --source-postgres-url <SOURCE_POSTGRES_URL> --target-postgres-url <TARGET_POSTGRES_URL>

Options:
      --bucket-name <BUCKET_NAME>
          S3 Bucket name where the CDC files are stored
      --s3-prefix <S3_PREFIX>
          S3 Prefix where the files are stored Example: data/landing/rds/mydb
      --source-postgres-url <SOURCE_POSTGRES_URL>
          Url of the database to validate the CDC files Example: postgres://postgres:postgres@localhost:5432/mydb
      --target-postgres-url <TARGET_POSTGRES_URL>
          Url of the target database to import the parquet files Example: postgres://postgres:postgres@localhost:5432/mydb
      --database-schema <DATABASE_SCHEMA>
          Schema of database to validate against S3 files [default: public]
      --included-tables [<INCLUDED_TABLES>...]
          List of tables to include for validatation against S3 files
      --excluded-tables [<EXCLUDED_TABLES>...]
          List of tables to exclude for validatation against S3 files
      --mode <MODE>
          Mode to load Parquet files Example: DateAware Example: AbsolutePath Example: FullLoadOnly [default: date-aware] [possible values: date-aware, absolute-path, full-load-only]
      --start-date <START_DATE>
          Start date to filter the Parquet files Example: 2024-02-14T10:00:00Z
      --stop-date <STOP_DATE>
          Stop date to filter the Parquet files Example: 2024-02-14T10:00:00Z
      --chunk-size <CHUNK_SIZE>
          Datadiff chunk size [default: 1000]
      --max-connections <MAX_CONNECTIONS>
          Maximum connection pool size [default: 100]
      --start-position <START_POSITION>
          Datadiff start position [default: 0]
      --only-datadiff
          Run only the datadiff
      --only-snapshot
          Take only a snapshot from S3 to target DB
      --accept-invalid-certs-first-db
          Accept invalid TLS certificates for the first database
      --accept-invalid-certs-second-db
          Accept invalid TLS certificates for the second database
  -h, --help
          Print help
  -V, --version
          Print version

Using Inquire

rust-cdc-validator --features="with-inquire"

Installation (Library)

In order to use the tool as a library, you can run:

cargo add rust-cdc-validator

Example

  • Spin up the local Postgres DB through Docker Compose:
docker-compose up

psql -h localhost -p 5438 -U postgres -d mydb
  • Build and run the Rust tool
cargo fmt --all
cargo clippy --all

cargo build

RUST_LOG=dms_cdc_operator=info,rust_pgdatadiff=info cargo run --features="with-clap" validate --bucket-name my-bucket --s3-prefix prefix/path --source-postgres-url postgres://postgres:postgres@localhost:5432/mydb1 --target-postgres-url postgres://postgres:postgres@localhost:5438/mydb --database-schema public --included-tables mytable --start-date 2024-02-14T10:00:00Z --chunk-size 100

For more debugging, you can enable Rust related logs by exporting the following:

export RUST_LOG=dms_cdc_operator=debug,rust_pgdatadiff=debug

License

This project is licensed under the MIT License

Dependencies

~91MB
~2M SLoC