fof / redis
Adds Redis cache, sessions and queue to Flarum
Requires
- flarum/core: ^2.0.0
- illuminate/redis: ^13.0
- predis/predis: ^v2.3.0
Requires (Dev)
- flarum/phpstan: 2.x-dev
This package is auto-updated.
Last update: 2026-07-28 16:10:57 UTC
README
This library allows using Redis as cache, session, settings storage, and for the queue. You can only
enable these services by using a local extender (the extend.php in
the root of your Flarum installation). See the "Set up" section below.
This is an advanced utility for webmasters!
Installation
Install manually with composer:
composer require fof/redis:"*"
Set up
In your extend.php:
return [ new FoF\Redis\Extend\Redis([ 'host' => '127.0.0.1', 'password' => null, 'port' => 6379, 'database' => 1 ]) ];
This enables sessions, cache, settings, and queue to run on redis.
phpredis vs Predis
If the PHP redis extension (ext-redis) is installed, it will be used automatically. Otherwise Predis is used as a fallback. No configuration change is required.
Persistent connections (recommended for phpredis)
With phpredis, connections can be reused across requests within the same PHP-FPM worker process. Add persistent and persistent_id to your config to enable this:
return [ new FoF\Redis\Extend\Redis([ 'host' => '127.0.0.1', 'password' => null, 'port' => 6379, 'database' => 1, 'persistent' => true, 'persistent_id' => 'flarum', // groups connections into a named pool per worker ]) ];
This avoids opening a new TCP connection on every request, which is a significant throughput improvement at scale. persistent_id is a string used to key the connection slot — use the same value across all workers on the same host.
Unix socket
return [ new FoF\Redis\Extend\Redis([ 'host' => '/var/run/redis/redis.sock', 'port' => 0, 'database' => 1, 'persistent' => true, 'persistent_id' => 'flarum', ]) ];
Other phpredis options
| Key | Example | Notes |
|---|---|---|
timeout |
2.0 |
Connect timeout in seconds |
read_timeout |
2.0 |
Per-command timeout |
retry_interval |
100 |
ms between reconnect attempts |
prefix |
'flarum_' |
Key prefix |
compression |
Redis::COMPRESSION_LZ4 |
phpredis ≥ 5.3, requires lz4/zstd compiled into ext-redis |
Multi-instance deployments: This extension can handle distributed cache invalidation across multiple Flarum instances (pods/containers) via Redis Pub/Sub. See DISTRIBUTED_CACHE.md for details.
Settings Cache
This extension includes a settings cache that significantly improves performance by caching all Flarum settings in Redis. This eliminates hundreds to thousands of database queries per page load.
How it works:
- All settings are cached in Redis with a dedicated connection (database 4 by default)
- Settings are loaded from cache on first access, then served from memory
- Cache is invalidated on any settings write, ensuring consistency across multiple instances
- Perfect for multi-container/multi-pod deployments where settings changes must propagate immediately
Configuration:
The settings cache uses database 4 by default. To customize:
return [ (new FoF\Redis\Extend\Redis([ 'host' => '127.0.0.1', 'password' => null, 'port' => 6379, 'database' => 1, ])) ->useDatabaseWith('cache', 1) ->useDatabaseWith('queue', 2) ->useDatabaseWith('session', 3) ->useDatabaseWith('settings', 4) // Settings cache database ];
To use a completely separate Redis instance for settings:
return [ (new FoF\Redis\Extend\Redis([ 'connections' => [ 'cache' => [ 'host' => 'cache.yoursite.com', 'database' => 1, ], 'settings' => [ 'host' => 'settings.yoursite.com', 'database' => 4, ], // ... other connections ], ])) ];
Performance impact:
- Reduces settings-related database queries by 97-99%
- Typical Flarum page load: 1,500+ settings queries → ~10 queries
- Cache invalidation ensures consistency in multi-instance environments
See "Use different database for each service" below to split up the database for cache vs sessions, queue because a cache clear action will clear sessions and queue jobs as well if they share the same database.
Advanced configuration
- Disable specific services:
return [ (new FoF\Redis\Extend\Redis([ 'host' => '127.0.0.1', 'password' => null, 'port' => 6379, 'database' => 1, ]))->disable(['cache', 'queue', 'settings']) ];
- Use different database for each service:
return [ (new FoF\Redis\Extend\Redis([ 'host' => '127.0.0.1', 'password' => null, 'port' => 6379, 'database' => 1, ])) ->useDatabaseWith('cache', 1) ->useDatabaseWith('queue', 2) ->useDatabaseWith('session', 3) ->useDatabaseWith('settings', 4) ];
- Completely separate the config array:
return [ (new FoF\Redis\Extend\Redis([ 'connections' => [ 'cache' => [ 'host' => 'cache.int.yoursite.com', 'password' => 'foo-bar', 'port' => 6379, 'database' => 1, ], 'queue' => [ 'host' => 'queue.int.yoursite.com', 'password' => 'foo-bar', 'port' => 6379, 'database' => 1, ], 'session' => [ 'host' => 'session.int.yoursite.com', 'password' => 'foo-bar', 'port' => 6379, 'database' => 1, ], 'settings' => [ 'host' => 'settings.int.yoursite.com', 'password' => 'foo-bar', 'port' => 6379, 'database' => 4, ], ], ])) ];
- Use a cluster:
return [ (new FoF\Redis\Extend\Redis([ 'host' => '127.0.0.1', 'password' => null, 'port' => 6379, 'database' => 1, 'options' => [ 'replication' => 'sentinel', 'service' => 'mymaster:26379', ] ])) ->useDatabaseWith('cache', 1) ->useDatabaseWith('queue', 2) ->useDatabaseWith('session', 3) ->useDatabaseWith('settings', 4) ];
Queue
Make sure to start your queue workers, see
the laravel documentation for specifics.
To test the worker can start use php flarum queue:work.
Queue options
The queue allows for several options to be added, retry_after, block_for and after_commit. You can set these
by adding a queue array in the configuration:
return [ (new FoF\Redis\Extend\Redis([ 'host' => '127.0.0.1', 'password' => null, 'port' => 6379, 'database' => 1, 'queue' => [ 'retry_after' => 120, // seconds 'block_for' => 5, // seconds 'after_commit' => true, 'failed_ttl' => 604800, // seconds; how long to keep failed jobs (default 7 days) 'queues' => ['default'], // additional named queues (see below) ] ])) ->useDatabaseWith('cache', 1) ->useDatabaseWith('queue', 2) ->useDatabaseWith('session', 3) ];
You can read up on the meaning of retry_after, block_for and after_commit in the
Laravel Documentation.
Named queues
Like Laravel Horizon, the Redis queue supports multiple named queues, which let you separate and
prioritise work — for example a fast notifications queue that should never be held up behind slow
exports.
Jobs are routed to a queue by name. A job can set its own target with the core
AbstractJob::$onQueue property (or Laravel's onQueue() method):
class SendExportJob extends \Flarum\Queue\AbstractJob { public static ?string $onQueue = 'exports'; }
You then run a worker across the queues you care about, in priority order — the worker fully drains each queue before moving to the next:
php flarum queue:work --queue=notifications,default,exports
Because Redis has no efficient way to list every queue that exists, Flarum keeps a registry of the queue
names admin tooling should know about (the queue dashboard and per-queue pause read it). Declare the extra
queues your site uses in the queues config key so they are covered:
'queue' => [ 'queues' => ['notifications', 'exports'], ],
default is always included automatically, so you only need to list the additional names. Queues you
don't declare still work — jobs pushed to them are processed normally — but they won't appear in the
dashboard or be individually pausable.
Failed jobs
When the queue runs on Redis, failed jobs are stored in Redis rather than in the database. Flarum
core only records failed jobs in the database for the database queue driver; with this extension enabled,
your failures stay on Redis so they don't add load or writes to your database. They remain fully visible
and manageable from the admin queue dashboard (view, retry, and delete), and via the
php flarum queue:failed, queue:retry and queue:forget commands.
Failed jobs are given a time-to-live so they don't accumulate forever, controlled by failed_ttl (seconds)
in the queue config above. It defaults to 7 days (604800). Set it to 0 or null to keep failed
jobs until they are retried or deleted. Because these entries carry a TTL, they are also eligible for
eviction first under a volatile-* maxmemory-policy, so a memory-constrained Redis reclaims old failure
records before touching live queue jobs.
Migrating from blomstra/flarum-redis
Simply update the namespace used in your extend.php file from Blomstra\Redis... to FoF\Redis...
Updating
composer update fof/redis
FAQ
Why are there still files in storage/cache? Some code still relies on physical files being present. This includes the formatter cache and the view caches.