All queries (including simple ones) are taking a min of 4-5 seconds to complete. How can I improve this?

[drewgallagher]

I tried to test the performance of a small query to see if anything is slowing down all requests built into the framework.

I tried a simple user query and logged the execution time of the actual query.

The query itself took very little time at all

[2023-06-01 13:30:14] local.DEBUG: App\Services\Slack\SlackGuzzleClient: The $auth_token argument is empty. The client will not be able to authenticate with Slack.  
[2023-06-01 13:30:14] local.DEBUG: App\Services\Slack\NovaSlackbot: The channel argument is empty. Messages will not be able to be posted to Slack.  
[2023-06-01 13:30:14] local.EMERGENCY: Query time: 0.0017249584197998  
[2023-06-01 13:31:34] local.DEBUG: App\Services\Slack\SlackGuzzleClient: The $auth_token argument is empty. The client will not be able to authenticate with Slack.  
[2023-06-01 13:31:34] local.DEBUG: App\Services\Slack\NovaSlackbot: The channel argument is empty. Messages will not be able to be posted to Slack.  
[2023-06-01 13:31:34] local.EMERGENCY: Query time: 0.0018429756164551  
[2023-06-01 13:31:39] local.DEBUG: App\Services\Slack\SlackGuzzleClient: The $auth_token argument is empty. The client will not be able to authenticate with Slack.  
[2023-06-01 13:31:39] local.DEBUG: App\Services\Slack\NovaSlackbot: The channel argument is empty. Messages will not be able to be posted to Slack.  
[2023-06-01 13:31:39] local.EMERGENCY: Query time: 0.0015189647674561  
[2023-06-01 13:31:44] local.DEBUG: App\Services\Slack\SlackGuzzleClient: The $auth_token argument is empty. The client will not be able to authenticate with Slack.  
[2023-06-01 13:31:44] local.DEBUG: App\Services\Slack\NovaSlackbot: The channel argument is empty. Messages will not be able to be posted to Slack.  
[2023-06-01 13:31:44] local.EMERGENCY: Query time: 0.0017650127410889  

However the insomnia request takes much longer as seen in the attached photo. On average it's around ~4-5 seconds.
Is there something with rebing/graphql or another integration that is adding overhead, slowing down our requests?

The code I used to test the query was here

        $start = microtime(true);
        LogService::save(\Request::post("query"), "info", "query");

        $user = User::findOrFail(Auth::id());
        $end = microtime(true);

        $time = $end - $start;
        Log::emergency("Query time: " . $time);
        return $user;

Below is the average execution time of that request via insomnia

[mfn]

These are certainly runtimes I would not expect, if not much happens in the resolvers. Of course there are a lot of variables, so it's hard to say.

Is your schema big? Lots of types?

Besides that, I suggest to add further debug information; maybe also use custom middlewares for this or directly edit vendor files temporarily to get better measurements. Or use some external tooling, I've never used blackfire but I've seen people with great success finding bottlenecks.

[drewgallagher]

We have around ~ 100 Types and and 234 tables

[mfn]

I just checked on a private project, > 200 types, > 100 queries, > 100 mutations and couple of other stuff registered. No bottlebeck known, also not in local dev an certainly not > seconds.

[drewgallagher]

Hm then maybe it's another library or middleware at is causing it. Maybe some auth middleware?

[drewgallagher]

@mfn do you have any profilers that you use? I'm trying to use Laravel debugbar, but it doesn't pop up when I hit the application port. Maybe it's because our app it's not using the regular built in php server?

[mfn]

See my first reply #1039 (comment)

I've never used blackfire but I've seen people with great success finding bottlenecks.

Maybe it's because our app it's not using the regular built in php server?

I'm using php-fpm (via nginx).

xdebug has profiling capability, too; you need further tools to "read" the profiling output (PhpStorm supports it).