Performance Issues with Laravel Application on Koyeb - Slow Page Load Times

Troubleshooting and help
1

We are experiencing significant performance issues with our Laravel application deployed on Koyeb. Looking to see if others in the community have faced similar issues or found solutions.

Environment Details:

  • Koyeb Web Service (Instance: CPU Standard Medium)
  • Koyeb Database (Instance: Medium)
  • Deployment: Using Buildpack
  • Laravel Version: 10.x
  • PHP Version: 8.2
  • UI Enhancement Library: [backpack]

Current Issues:

  • Initial page loads and page transitions take 5-10 seconds
  • Disk I/O appears to be unusually slow
  • These issues persist despite running on Linux + Docker environment

What We’ve Investigated:

  1. Tested using Koyeb persistent volume for /var/www/html/vendor with a simpler Laravel application
  • Performance improvements were observed
  • However, operational challenges exist (inability to unmount volumes, additional steps required for cache clearing)
  1. Aware of common vendor directory mounting issues with Docker + Laravel
  • These should not have significant impact in Linux environments

Questions:

  1. Are there any Koyeb-specific configurations available to improve I/O performance?
  • Specifically regarding disk I/O configuration options
  • Possibilities for filesystem performance tuning
  1. Regarding the use of persistent volumes for vendor directory:
  • While performance improvements were confirmed, we face these operational challenges:
    • Unable to unmount volumes
    • Need for cache clearing with fixed volumes
    • Increased operational costs
  • Are there any alternative solutions to these challenges?
  1. Regarding AWS backbone usage:
  • When selecting AWS as the backbone for Koyeb web service, does it effectively become an Amazon Linux 2 + Docker configuration?
  • Could this configuration change potentially resolve our current performance issues?

Additional Question:
4. Community Experience:

  • Has anyone else in the Koyeb community experienced similar performance issues with Laravel applications?
  • If so, what solutions or workarounds have you found effective?
  • Are there any known issues or discussions about Laravel performance on Koyeb’s platform?

Thank you for your support.

2

Thanks for the detailled report!

On which region have you been running your tests?

All our servers are backed by high-performance NVMe drives, so disk I/O shouldn’t be an issue.
Deploying on the AWS region will not give you better performance, something sounds abnormal.

Do you have an easy way for us to reproduce your issue?

3

Thank you for your response.

I tested mounting the /var/www/html/vendor directory as a fixed volume in a container running on Koyeb with PHP + Laravel + Backpack, and I observed a noticeable improvement in response speed.

For lightweight applications, the performance is inherently good, making it harder to confirm improvements. However, I believe testing is necessary for applications with relatively heavy implementations, such as the Laravel application for Backpack that I am using, where the libraries under /var/www/html/vendor play a significant role.

I apologize for the inconvenience, but I would greatly appreciate it if you could confirm this behavior.

For reference:

  • On a local Linux (Ubuntu) environment with Docker running a Laravel application, this issue does not occur. It seems Docker’s disk access does not introduce overhead in this setup.
  • On a Windows environment with Docker running a Laravel application, the issue does occur. Mounting the /var/www/html/vendor directory as a separate fixed Docker volume significantly improved performance.

Note: The issue of improving performance for the /var/www/html/vendor directory in Laravel + Docker setups is widely discussed online and may be worth researching.

Since this is a critical performance issue related to the operating environment, we are in great need of a resolution. Your cooperation in addressing this would be highly appreciated.

Let me know if you’d like further revisions!

4

Additionally:

  • The region being used is Tokyo.
  • The instance size is CPU Standard Medium.
  • According to the metrics, CPU usage remains at 0% with no visible CPU load on the graph.
  • As for memory, the total available is 2GB, and only about 250MB is being utilized.
5

Hi @terukazu_itou,

Is there any chance you could share access to your repo with us?

We would need to reproduce the issue to investigate.