horizon:clear-metrics does not properly clear metrics

[mikescott]

Horizon Version

5.45.4

Laravel Version

12.55.1

PHP Version

8.4

Redis Driver

PhpRedis

Redis Version

6.3.0

Database Driver & Version

No response

Description

I enabled metrics in Horizon as per the instructions after several months in production. The first node on the chart showed a count of over 30,000 which obviously distorted the scale of the chart massive.

So I ran php artisan horizon:clear-metrics which cleared the contents of the pages /horizon/metrics/jobs and /horizon/metrics/queues

But when a new job came in the metrics came back, the new chart starts at 30,000 again.

Also after clearing metrics, and before any new job has run the page at /horizon/metrics/jobs/App\Path\To\Some\Job still has a chart with data even though it's no longer listed at /horizon/metrics/jobs

Steps To Reproduce

Run php artisan horizon:clear-metrics

Metrics are not properly cleared.

[mikescott]

In Horizon.php on line 109:

  $config['options']['prefix'] = config('horizon.prefix') ?: 'horizon:';

In RedisMetricsRepository.php on line 399:

  $scanResult = $this->connection()->scan(
      $cursor ?? 0, ['match' => config('horizon.prefix').$pattern]
  );

The issue appears as though it's because the match pattern passed to scan() is double-prefixed, meaning nothing gets deleted.

[github-actions]

Thank you for reporting this issue!

As Laravel is an open source project, we rely on the community to help us diagnose and fix issues as it is not possible to research and fix every issue reported to us via GitHub.

If possible, please make a pull request fixing the issue you have described, along with corresponding tests. All pull requests are promptly reviewed by the Laravel team.

Thank you!

[webpatser]

Hi @mikescott — I investigated this and wanted to share what I found, since it might help narrow down the root cause.

The current clear() code works correctly in the default PhpRedis configuration. I wrote a test that creates metrics, takes a snapshot, calls clear(), and verifies everything is empty — it passes. The scan pattern config('horizon.prefix').$pattern correctly matches keys when OPT_SCAN is at its default value (0).

The "double-prefix" theory only holds when PhpRedis has SCAN_PREFIX enabled (via Redis::OPT_SCAN). With SCAN_PREFIX, PhpRedis auto-prepends the connection prefix to scan match patterns, so manually adding it again causes a double-prefix (horizon:horizon:snapshot:*). This is what PR #1752 tried to fix, but it was closed because the issue description doesn't mention SCAN_PREFIX.

Could you check your Redis configuration?

1. Do you have 'scan' => Redis::SCAN_PREFIX (or Redis::SCAN_RETRY | Redis::SCAN_PREFIX) set in your database.php Redis options?
2. What does your config/horizon.php prefix value look like?
3. Do you have a global Redis prefix set in database.php → options.prefix in addition to Horizon's prefix?

That would help pinpoint exactly which configuration triggers this.