AWS 上的 WordPress，带有 ECS、EFS 和 ELB 返回 502

Question

I have a site based on WordPress 4.9.6 that is deployed using AWS ECS, with EFS used to store WordPress' files. As database I use an AWS Aurora MySQL 5.7 compatible instance. I have also set up an application load balancer to access the containerized WordPress instances. (More specifics on the setup below.)

Problem overview

This setup does seem to work in most cases. Ie I am able to do GET requests on the site. I can login in, see the dashboard and often times do updates successfully. The problem I face is that also often times my update attempts result in 502 Bad Gateway response when I commit my update, ie doing POST /wp-admin/post.php .

Specifics

Setup

First my DB instance writer is populated with a dump of my local dev database. The database URL entries pointing back to the site itself, eg siteurl and home in the options table, have values with https .

The EFS was created with performance mode General Purpose . I originally tried with mode Max I/O , but resources suggesting to rather use the former mode. However, toggling performance mode has not change the frequency of 502 errors.

My Amazon Linux ECS cluster instances are provisioned to mount an EFS volume using NFSv4.1 as suggested by AWS ( mount options nfsvers=4.1,rsize=1048576,wsize=1048576,hard,timeo=600,retrans=2 ).

I have a Docker image derived from the official WordPress image in which I update the original image /usr/src/wordpress with my custom content. The custom content includes ao an updated wp_config.php in which I set $_SERVER['HTTPS']='on';

Putting pieces together my ECS task definition uses my custom Docker image and mounts a directory on my EFS drive as Docker volume in the Docker container. Bottom line here is that when I create an ECS service with the task definition the container spins up and writes to /var/www/html which I subsequently can see in my EFS drive. This all seems fine and dandy.

My containerized WordPress instances are then successfully registered in a target group I've previously set up for an application load balancer.

I can then access my site over https protocol. If I try to use http I am redirected to https as planned. I can open the landing page. When I log in I try to edit the landing page.

Problem

This is where I face the real problem. Often times, but not always, when I do an update on the landing page and click the Update button I get a 502 Bad Gateway response to the POST /wp-admin/post.php request. Also often when I start to edit and GET /wp-admin/post.php?post=2&action=edit is requested I get the same 502.

I don't see much pattern in when I do and do not get the 502. I have tried to update both textual contents as well as adding images to the landing page. The 502 happens sometimes but not always in either case.

Also I tried to remedy the problem as I suspected it had to do with the use of EFS and subsequent synchronization problems between the two ECS instances I have set up for the test. The following attempts were made, without seeing significant improvement.

• Add mount option sync as suggested by the EFS user guide in order to avoid caching on the ECS instance
• Increase the Idle timeout of the load balancer

Finally I reduced the number of ECS service tasks from 2 to 1 but the problem still persists.

As an error message in the browser console when meeting the 502 I often, possibly always, see the following: The character encoding of the HTML document was not declared. The document will render with garbled text in some browser configurations if the document contains characters from outside the US-ASCII range. The character encoding of the page must be declared in the document or in the transfer protocol. The character encoding of the HTML document was not declared. The document will render with garbled text in some browser configurations if the document contains characters from outside the US-ASCII range. The character encoding of the page must be declared in the document or in the transfer protocol.

So, does anyone have clue about what to try next and where to look for indications to the reason for the problem?

Answer 1

It looks like following this

https://wordpress.org/support/article/editing-wp-config-php/#increasing-memory-allocated-to-php

Helped resolve my issue, the usual time period where it would time out it has stopped. TLDR; During the Dockerfile build process, I added the following lines

# Copy php.ini Over to Container
COPY php.ini /usr/local/etc/php/
COPY phpinfo.php /var/www/html/wp-admin/
COPY .htaccess /var/www/html

Php.ini

upload_max_filesize = 500M
post_max_size = 256M
memory_limit = 256M
max_execution_time = 1800
max_input_time = 180
max_input_vars = 5000

.htaccess

php_value upload_max_file 500M
php_value post_max_size 256M
php_value memory_limit 256M
    

phpinfo.php This is accessible by navigating to /wp-admin/phpinfo.php It will give you a summary of the current WP Configuration to see if the changes have been implemented or not, also great for troubleshooting

<?php
phpinfo();
?>

You could either update your task definitions with the new Container built into the ECR, (Not sure if you've automated this ? )

Or just access the EC2's via Bastion host and then docker exec -it and edit the files manually, Seeming that you have an EFS it'll be static.

This helped my time out issue. Let me know how you go!