nginx: fallback to try_files when proxy_pass fails requires unusual config - nginx

I am using Nginx to server a single page app. Basically we just need to serve the index.html page whenever no matching file is found. The location looks like this and has been working just fine:
location / {
try_files $uri $uri/ /index.html
}
Now I would like to query an upstream server, and only if that fails, use the try_files directive as above
If the try_files is just moved to a fallback location like
location #fallback {
try_files $uri $uri/ /index.html;
}
location / {
proxy_pass http://127.0.0.1:8080;
proxy_intercept_errors on;
error_page 400 403 502 503 504 #fallback;
}
then - when the upstream server is unavailable - the client sees the Nginx 502 error page instead of the files served from the file system.
I finally found a solution that works by using a double slash in front of the /index.html fallback. This is the whole config file which can be used with the official nginx docker image for testing
events {
}
http {
error_log /var/log/nginx/error.log;
access_log /var/log/nginx/access.log;
server {
listen 80;
root /usr/share/nginx/html/;
location / {
proxy_pass http://127.0.0.1:9990;
proxy_intercept_errors on;
error_page 400 403 502 503 504 = #fallback;
}
location #fallback {
try_files $uri?$args /index.html //index.html;
}
}
}
which can be run with a command like
docker run -v /path/to/www/folder:/usr/share/nginx/html:ro -v /path/to/config/nginx.conf:/etc/nginx/nginx.conf -d -p 8080:80 nginx
In case no double slash is present before the last index.html fallback like
location #fallback {
try_files $uri?$args /index.html;
}
Then nginx constructs a path on the filesystem like <root>index.html, which has a missing delimiter instead of the correct <root>/index.html, whenever a url which is not the root url is requested.
Final question: why does this setup require a double slash within the try_files directive? Why can't one just use the try_files section from a regular config and move it to a fallback location used when intercepting errors?

I was presented with a similar situation, and I solved this going the other way around, using the suggestion from this page of common pitfalls. That is, first serve the static files, and then fallback to the proxy:
location / {
try_files $uri $uri/ #proxy;
}
location #proxy {
proxy_pass http://127.0.0.1:9990;
}
In this case, this would first look for the presence of the files as static files in the root, then proxy the request to http://127.0.0.1:9000. This is functionnally equivalent unless you want the files from the proxy to shadow the static files.

Related

Site is not redirecting to backend node NGINX config

I have a react frontend with a node backend but my nginx setup is currently not working. What I want to do is redirect https://sales.example.com/api/stats/get_customer_count/2323232 to http://127.0.0.1:3000/stats/get_customer_count/2323232(Node backend server runs on http://127.0.0.1:3000) but I keep getting 404 errors stating that the api path is not found. Not sure how to go about fixing this. Appreciate it. Thanks
server {
root "/home/sales/frontend/dist";
index index.html index.htm;
server_name sales.example.com;
error_log /var/log/nginx/error.log;
listen 80;
location / {
index index.html index.htm;
try_files $uri $uri/ /index.html =404;
}
location ~* ^/api/ {
rewrite ^/api/(.*) /$1 break;
proxy_pass http://127.0.0.1:3000;
}
}
This seems to be answered here already, and the explanation written by #Dayo is good.
Translated it to your example, it looks like this: (notice, the main difference is the tailing slash in the proxy pass)
location ~* ^/api/ {
proxy_pass http://127.0.0.1:3000/;
}
But please, read through the linked answer, before copying this over.

Reverse proxy nginx to itself

I am currently hosting a single-page react app that is hosted in the URL root like so:
server {
listen 80;
server_name localhost;
location / {
root /var/www/html;
try_files $uri /index.html;
}
}
I need to put the site behind an AWS elastic load balancer and at the same time change the path so everything is within a /support directory e.g. http://example.com/index.html -> http://example.com/support/index.html.
AWS ALBs do not support URL rewriting so I have to do this within the nginx config on the server. First of all I tried changing the config to:
server {
listen 80;
server_name localhost;
location /support {
alias /var/www/html;
try_files $uri /index.html;
}
}
This sort-of works but the URLs within the javascript content don't contain the /support path (e.g. they contain http://example.com/script.js instead of http://example.com/support/script.js).
I then tried creating a reverse-proxy config to proxy /support to /, which sadly put nginx in an infinite loop until it ran out of worker threads:
server {
listen 80;
server_name localhost;
location /support {
proxy_pass http://localhost:80;
}
location / {
root /var/www/html;
try_files $uri /index.html;
}
}
I'm confused why requests are going into a reverse-proxy loop? Shouldn't proxy_pass remove the /support prefix before proxying the request, and therefore it shouldn't be "caught" again by the /support location?
Just a guess.
Do you want to serve something on /?
If not - it is easy:
server
{
listen 80;
server_name localhost;
location /support/
{
alias /var/www/html/;
try_files $uri $uri/ /index.html;
}
location /
{
return 303 http://localhost/support$request_uri;
}
}
Fiddle around with the ending slashes if it does not work (using them - or not - makes often a difference).
Use alias instead of root so that /support is not added to the /var/www/html folder.
Everything gets redirected to /support.
If you want to serve something on / which is different from /support:
Use sub_filter or subs_filter in /support to rewrite your source code links on-the-fly so that they will never use /.
If you have redirects inside your source code (or proxy_pass backend) - you need proxy_redirect and/or Lua to catch and change them on-the-fly.

Dokku redirects to another domain when requested site is down

I have Dokku installed on a server, with multiple sites/domains deployed to it. When one of my sites goes down, all HTTP requests to it get redirected (for some reason) to another site. This is confusing. I'm expecting Dokku to show some error page in this case. Is it the default behavior or I did something wrong?
PS. This is the problem: https://github.com/dokku/dokku/issues/2602
How about adding a custom error page based on the error code by editing vhost file:
server{
server_name www.foo.com;
root /srv/www/foo/public_html;
expires 1M;
access_log /srv/www/foo/logs/access.log;
error_log /srv/www/foo/logs/error.log;
error_page 404 /404.html;
location / {
index index.html;
rewrite ^/(.*)/$ /$1 permanent;
try_files "${uri}.html" $uri $uri/ =404;
}
location = /404.html {
internal;
}
}
Your server error might be caught from codes 404 or 500

Using try_files with uwsgi

I am trying to use the nginx try_files directive with uwsgi_pass and having a ton of difficulty.
Basically what I want is for try_files to ask the uWSGI container if the request URI is valid and if not, then serve up the index.html file instead. My nginx config is as follows:
server {
listen 80;
access_log /tmp/nginx.log;
location / {
try_files $uri $uri/ /index.html;
include uwsgi_params;
uwsgi_pass 127.0.0.1:5001;
}
}
But what this does is check the docroot for every request and if its not there, it simply bails and returns the index.html file.
What I want instead is the following:
Request comes in for www.myapp.com
nginx forwards this request onto the uWSGI container
If that is invalid, then return the index.html
Is there a way to 'ask' uWSGI to try the files instead?
What I'm ultimately trying to accomplish here is HTML5 Pushstate with React Router. I'm running a Flask app with a React front-end. If the user refreshes the browser at www.myapp.com/preferences/userid, then I want nginx to forward that to the container and if its invalid, to return the index.
So, after talking with #Chamindu, I realized I was probably going about this the wrong way. I prevented uWSGI from serving my index.html (even though it could) and instead relied on nginx to serve that instead.
server {
listen 80;
access_log /tmp/nginx.log;
location / {
root /var/www/myapplication/;
try_files $uri $uri/ /index.html;
}
location /api {
include uwsgi_params;
uwsgi_pass 127.0.0.1:5001;
}
}

403 forbidden on wordpress index with nginx, the rest of the pages work fine

I'm setting up my blog on a new EC2 instance because one of the sites on the server that's currently hosting it is being DDoSed.
I'm having some trouble with nginx, because I can either see all the pages fine but 403 on the index, or see the index but 404 on the pages (depending on the config I'm using)
Here's my nginx config:
server {
listen 80;
server_name www.test.com;
server_name test.com;
root /www/blog;
include conf.d/wordpress/simple.conf;
}
And simple.conf:
location = /favicon.ico {
log_not_found off;
access_log off;
}
location = /robots.txt {
allow all;
log_not_found off;
access_log off;
}
location / {
# This is cool because no php is touched for static content.
# include the "?$args" part so non-default permalinks doesn't break when using query string
try_files $uri $uri/ /index.php?$args;
}
location ~ \.php$ {
#NOTE: You should have "cgi.fix_pathinfo = 0;" in php.ini
include fastcgi.conf;
fastcgi_intercept_errors on;
fastcgi_pass unix:/var/run/php-fpm/php-fpm.sock;
}
location ~* \.(js|css|png|jpg|jpeg|gif|ico)$ {
expires max;
log_not_found off;
}
if I change the try_files $uri $uri/ /index.php?$args; to index index.php, the front page will work fine and the rest will be 404. If I leave it like that, the front page is 403.
Here's the error log:
2013/08/07 19:19:41 [error] 25333#0: *1 directory index of "/www/blog/" is forbidden, client: 64.129.X.X, server: test.com, request: "GET / HTTP/1.1", host: "www.test.com"
That directory is 755 on the nginx user:
drwxr-xr-x 6 nginx nginx 4096 Aug 7 18:42 blog
Is there anything obvious I'm doing wrong ?
Thanks !
Add index index.php; In the server block, if it doesn't work then you need to remove the $uri/ because you don't want to do a autoindex on
EDIT: Just noticed that you already figured out your problem, so I'll add the reasoning behind it, the reason why you needed autoindex on; is because without it nginx will follow the try_files rules,
Check if there's a file called /, and of course it fails.
Check if there's a directory called / (by adding root it would = /www/blog/), this check will succeed, so it tries to list the content of the folder.
Since you didn't specify autoindex on; so by default nginx should forbid directory listing, thus it would return a 403 forbidden error.
The rest of the site works fine because it fails the $uri/ test or doesn't reach it, because you probably don't have a folder called image.jpg or stylesheet.css etc.
Looks like I needed the inded index.php in the server {} definition and not in the location {}
It seems that you are not allowing arguments to be sent to the CMS so this will not show this uris that would bring information from the database and redirect you to the 403 page.

Resources