Checking your setup
views:doctor looks at the setup and says what to fix. Run it after installing, after changing the config and on the
servers that record views, since it reads their cache, scheduler and Redis:
php artisan views:doctor Shared cache stores ✗ Cooldowns: kept in the session, whose `array` driver forgets them after every request. → Use another session driver, or set `cooldown.store` to `cache`.
Scheduler ! `views:maintain` is not scheduled, so rollups, retention and counter columns are not kept up to date. → Add `Schedule::command('views:maintain')->hourly()->onOneServer();` to `routes/console.php`, unless a crontab of your own runs it.
ERROR 1 failure, 1 warning, 0 suggestions.| Check | Key | Looks at |
|---|---|---|
| Legacy config | legacy-config |
top-level keys of a v8 config, which v9 ignores, and a cooldown block without a store |
| Database schema | schema |
the views table, its columns and indexes, and the retention, rollup and counter columns in use |
| Optional indexes | index-advice |
the optional indexes the config relies on or a large table would use |
| Dimensions | dimensions |
that a capped dimension rollup has MySQL 8.0, that no personal dimension is folded, and how far back and how large each one’s rollup is |
| Rollups | rollups |
that every rollup counts the views it should, which anonymising and pruning wait for |
| Shared cache stores | shared-cache |
that cooldowns, the throttle, the burst guard, the fingerprint salt and remembered counts use a shared store |
| Scheduler | schedule |
that views:maintain or MaintainViewsJob, and views:flush, are scheduled when something needs them |
| Trusted proxies | trusted-proxies |
that a proxy is trusted when the fingerprint identity, IgnoreIpAddresses or IgnoreBursts reads the IP |
| Redis stream | redis-stream |
the server and its version, how long buffered views wait, and views a flush took without acknowledging |
| Refused attempts | crawler-share |
the share of attempts each guard refused, once sampling is on |
| Configuration | configuration |
settings that undo each other, such as queueing on the sync connection |
A failure means something does not work, a warning that it breaks under load or on more than one server, and a
suggestion that it could be better. The command fails on a failure, so it can run in a deploy. --strict fails on a
warning as well, --only=schema --only=schedule runs some of the checks and --json prints the findings for a script.
The console cannot see the queries an app runs or the requests it serves, so some checks read what the config implies.
The scheduler check reads the Laravel scheduler, so a command run from a crontab of your own shows as missing. The
proxies check reads the proxies trustProxies(at:) configures in bootstrap/app.php.
The share of attempts each guard refuses is only known while it is counted. Turn sampling on to count them per day in the cache for a week:
'doctor' => [ 'sample' => [ 'enabled' => true, 'store' => 'redis', 'crawler_share' => 0.5, ],],That adds two cache calls to every attempt, in a store every server must share. The doctor then warns when
IgnoreCrawlers refused more than crawler_share of the attempts, which usually means a client of your own, such as the
webview of an app, is taken for a crawler, or a scraper is crawling the site.