Configuring nginx on Google App Engine Flexible Environment with Gunicorn

nginx python google-cloud-platform google-app-engine

5,210

I believe there is no actual need for you to be touching Nginx. The link you are basing yourself on has to do with the HTTP Load Balancer, not with App Engine Flexible.

An example of working Django application on App Engine flexible may be found here.

5,210

Furkanzmc

Updated on September 18, 2022

Comments

Furkanzmc over 1 year
I'm working on a Django app and I was deploying it to App Engine Standard environment until now, but then I started using Google Cloud Vision Python library and due to some restrictions I ad to switch to Flexible Environment.

My code works correctly on my local machine with python manage.py runserver, but when I deploy it to App Engine I get the following
```
<html>
    <head>
        <title>502 Bad Gateway</title>
    </head>
    <body bgcolor="white">
        <center>
            <h1>502 Bad Gateway</h1>
        </center>
        <hr>
        <center>nginx</center>
    </body>
</html>
```
When I researched the problem, I found out that I need to change nginx settings as stated here.

Nginx Settings to Apply
```
# Tune nginx keepalives to work with the GCP HTTP(S) Load Balancer:
keepalive_timeout 650;
keepalive_requests 10000;
```
But I was unable to find a way to do this. I'm also new to backend development so I don't have much experience.

Here's my app.yaml file
```
runtime: python
api_version: 1
threadsafe: yes
env: flex
entrypoint: gunicorn -b :$PORT app.wsgi

runtime_config:
  python_version: 3

handlers:
  - url: /static
    static_dir: static
  - url: .*
    script: app.wsgi.application

builtins:
  - django_wsgi: on
```
I tried setting entrypoint to the following but it didn't work.
```
entrypoint: gunicorn --keep-alive 650 -b :$PORT app.wsgi
```
EDIT 1: This error only occurs with POST requests.

EDIT 2: Here's the error log with http request:
```
{
 insertId: "1c0for6fm6z812"   
 jsonPayload: {
  trace: "$traceId"    
  latencySeconds: "30.529"    
  time: null    
 }
 httpRequest: {
  requestMethod: "POST"    
  requestUrl: "/users/register/"    
  status: 502    
  responseSize: "568"    
  userAgent: "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/57.0.2987.133 Safari/537.36"    
  remoteIp: "130.211.3.116"    
  referer: "-"    
 }
 resource: {
  type: "gae_app"    
  labels: {
   project_id: "cuz-api"     
   version_id: "20170504t110352"     
   module_id: "default"     
  }
 }
 timestamp: "2017-05-04T08:30:14.560885079Z"   
 labels: {
  compute.googleapis.com/resource_name: "68f73c569efa"    
  compute.googleapis.com/resource_id: "8300604998554181598"    
  compute.googleapis.com/zone: "us-central1-b"    
  appengine.googleapis.com/trace_id: "7835424c5d67f87f56649adf26e71250"    
 }
 logName: "projects/cuz-api/logs/appengine.googleapis.com%2Fnginx.request"   
}
```
And this is the error log with https request:
```
{
 insertId: "95tl7qfm6m4zy"   
 jsonPayload: {
  time: null    
  trace: "$traceId"    
  latencySeconds: "30.028"    
 }
 httpRequest: {
  requestMethod: "POST"    
  requestUrl: "/users/register/"    
  status: 502    
  responseSize: "568"    
  userAgent: "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/57.0.2987.133 Safari/537.36"    
  remoteIp: "130.211.3.237"    
  referer: "-"    
 }
 resource: {
  type: "gae_app"    
  labels: {
   version_id: "20170504t110352"     
   module_id: "default"     
   project_id: "cuz-api"     
  }
 }
 timestamp: "2017-05-04T08:26:38.540802339Z"   
 labels: {
  compute.googleapis.com/zone: "us-central1-b"    
  appengine.googleapis.com/trace_id: "3eab23e9502e250d9f00dd5bc7eb465e"    
  compute.googleapis.com/resource_name: "68f73c569efa"    
  compute.googleapis.com/resource_id: "8300604998554181598"    
 }
 logName: "projects/cuz-api/logs/appengine.googleapis.com%2Fnginx.request"   
}
```
- Nick about 7 years
  
  Does the entrypoint command work to run your app if given PORT=8080 in your dev environment?
- Furkanzmc about 7 years
  
  I'm using manage.py runserver in my local development environment. The error occurs only with the POST requests. Apparently I forgot to mention that, I apologize. I'm editing my question to reflect that information.
- Nick about 7 years
  
  Well at this point the natural question to ask, and the natural test to perform, is: "what happens when you run gunicorn --keep-alive 650 -b :$PORT app.wsgi?"
- Furkanzmc about 7 years
  
  I get the same error.
- Nick about 7 years
  
  When running that command in dev, you send a POST request and get the exact 502 html payload shown in your question? And this is no differnt from when you run it using "manage.py"?
- Furkanzmc about 7 years
  
  The 502 error happens when I deploy the project to the Flex environment. I can run and send POST request without any problems on my local machine.
- Nick about 7 years
  
  You had said that running gunicorn in dev produces the same error. Do you not see the error in dev?
- Nick about 7 years
  
  Have you inspected all logs to ensure that there's no information on the error there?
- Nick about 7 years
  
  Finally, can you try to build and run the docker image on your local machine, to see if the issue is really in production or simply the image itself?
- Furkanzmc about 7 years
  
  I'm sorry, I said runnging python manage.py runserver doesn't give me the same error on my local machine. But I get the same error If I run the gunicorn command on my local machine. I checked the logs and the problem got me to this solution (As mentioned in my question). I just don't know how to apply that solution.
- Furkanzmc about 7 years
  
  I'm sorry. You are right. I added the error logs I got from GAE to my question. I also tried running gunicorn -b 127.0.0.1:8080 cuz.wsgi on a macOS, and I did NOT get the same error. It worked the way it's supposed to.
- Nick about 7 years
  
  It's not clear then that what you're seeing is actually related to the link in question then. At any rate the 502's only occur in production. There are other places you can check for logs, such as by SSH'ing into the actual instance itself, and once in there, you can also check inside the container running the app or the container running nginx.