README

This package is a Laravel wrapper for Foggy.

Configuration of the plugin can be found at Foggy's docs.

Install

Via Composer

composer require worksome/foggy-laravel

Usage

This package adds a new artisan command for running Foggy. Simply type:

php artisan db:dump

It will by default assume that the Foggy config file, will be in foggy.json in the root of the project.
A configuration file can be supplied by using --config argument.

The artisan command by default will make the database dump to stdout. To pass the output to a file, use the --output (-o) option.

php artisan db:dump --output scrubbed-dump.sql

Foggy also supports specifying a custom database connection:

php artisan db:dump --connection mysql

Uploading a dump to a filesystem disk

db:dump-to-disk streams the scrubbed dump through gzip and onto any configured filesystem disk. Nothing is written to local disk on the way, so it works in a container with no spare space — which starts to matter once a dump is tens of gigabytes. Its size is not known up front, so on S3 a dump of 5 MB or more goes up as a multipart upload and a smaller one as a single PutObject.

php artisan db:dump-to-disk --disk=dumps --connection=replica

It writes two objects:

dumps/dump-YYYY-MM-DD-HHMMSS-xxxxxx.sql.gz, the dump itself
dumps/latest, naming the most recent complete dump

Read the pointer rather than guessing a filename. Keys carry a timestamp and a random suffix, so a re-run never reuses the key of a published dump even when it starts in the same second, and latest is only written after the dump has finished and passed the scan below — so a truncated or rejected dump is never the one consumers pull.

Options: --disk, --prefix (default dumps), --name (default dump), plus --connection and --config, which are passed through to db:dump.

Checking that the rules actually worked

A Foggy config can only protect the columns it knows about. A new column on a table that is already withData: true is dumped raw, and neither the config nor the schema reveals that.

So the upload scans its own output, and fails without writing latest if it finds anything. A rejected dump fails the final read of the stream, so a disk that only stores an object once its stream ends cleanly — S3, for one — never stores it, and a write-only credential needs no delete permission to stay safe. One caveat: a dump large enough for a multipart upload leaves its uploaded parts behind, since the AWS SDK does not abort the upload — give the bucket an AbortIncompleteMultipartUpload lifecycle rule so they expire. The defaults look for email addresses, IBANs and Danish CPR numbers — deliberately a short, high-confidence list, because a scan that cries wolf gets switched off. Replace them per project:

// config/foggy.php
return [
    'scan_patterns' => [
        'email address' => '/[\w.%+-]+@(?!example\.(?:org|com|net)\b)[\w-]+\.[a-z]{2,}/i',
        'employee number' => '/\bEMP-\d{5}\b/',
    ],
];

The patterns are replaced, not merged, so a default that is noisy against your data can simply be dropped.

One constraint: the scan reads the dump in chunks and carries 256 bytes between them, so a pattern that can match more than that could miss an occurrence straddling a chunk boundary. The defaults are well inside it (an email address caps out at 254 characters). If you add something longer, widen the window:

// config/foggy.php
return [
    'scan_overlap_bytes' => 1024,
];

Patterns are validated when the scanner is constructed — one that will not compile raises InvalidArgumentException rather than silently matching nothing. A pattern that compiles but then fails mid-dump, such as a /u pattern meeting the non-UTF-8 bytes of a BLOB column, fails the run too: an unfinished scan is not evidence that the dump is clean.

worksome / foggy-laravel

Package info

Security

Statistics

README

Install

Usage

Uploading a dump to a filesystem disk

Checking that the rules actually worked