问题
I have a CentOS server running NGINX listening to 80 and a DB servering an app on 8080. I want to be able to Type
http://example.com/dev/abc
and have it actually access
http://example.com:8080/apex/abc or http://localhost:8080/apex/abc
I have used this location configuration
location /dev {
proxy_pass http://example.com:8080/apex;
}
However when I try it out the url displayed is
http://example.com/apex/apex
the page is not found and the log says:
2018/06/14 12:51:33 [error] 7209#0: *2067 open()
"/usr/share/nginx/html/apex/apex" failed (2: No such file or directory),
client: 124.157.113.187, server: _, request: "GET /apex/apex HTTP/1.1", host: "example.com"
Looks like two strange things are happening
1) Port 80 not 8080 is being used despite the proxy_pass
2) why is apex twice "/apex/apex/"
Help please :)
Adding entire Server block from config file:
server {
listen 80 default_server;
listen [::]:80 default_server;
server_name example.com;
root /usr/share/nginx/html;
# Load configuration files for the default server block.
include /etc/nginx/default.d/*.conf;
location /dev {
proxy_pass http://example.com:8080/apex;
}
error_page 404 /404.html;
location = /40x.html {
}
error_page 500 502 503 504 /50x.html;
location = /50x.html {
}
}
Update - More information on what the app that might help
The app is Oracle Application Express (APEX) it listens on port 8080. The URL works as follows:
HTTP://example.com:8080/apex/f?p=[APP]:[Page]:[Session] etc
Where [APP],[Page] and [Session] are all corrisponding numbers
The development environment url is actualy:
http://example.com:8080/apex/f?p=4550
This is the default so if I try http://example.com:8080/apex/ it defaults to http://example.com:8080/apex/f?p=4550 and takes you to the login page
Everything after the app number never changes so that is what I want to replaced by /dev/ http://example.com:8080/apex/f?p=4550:1 -> http://example.com/dev/:1
Once I have leant how this works, I plan to setup three proxy_pass's
example.com/dev -> http://example.com:8080/apex/f?p=4550
example.com/desktop -> http://example.com:8080/apex/f?p=1001
example.com/mobile -> http://example.com:8080/apex/f?p=201
Where the only thing that changes is the app number.
Rewrites are working fine for all three but I don't want the rewrite to be visible in the URL
Here are the rewrites:
location ~ /dev {
rewrite ^/dev(.*) http://smallblockpro.com:8080/apex$1 last;
}
location ~ /desktop/ {
rewrite ^/desktop/(.*) http://smallblockpro.com:8080/apex/f?p=1001:$1 last;
}
location ~ /desktop {
rewrite ^/desktop(.*) http://smallblockpro.com:8080/apex/f?p=1001:$1 last;
}
location ~ /mobile/ {
rewrite ^/mobile/(.*) http://smallblockpro.com:8080/apex/f?p=201:$1 last;
}
location ~ /mobile {
rewrite ^/mobile(.*) http://smallblockpro.com:8080/apex/f?p=201:$1 last;
}
回答1:
location ~ /desktop/ { rewrite ^/desktop/(.*) http://smallblockpro.com:8080/apex/f?p=1001:$1 last; }
The reason you're getting the :8080 port number shown up to the user is because you use absolute URLs in your rewrite directives, which results in NGINX producing 301 Moved responses directly to the user — your presumed expectation that it'll still go through proxy_pass after a rewrite like that is incorrect, see http://nginx.org/r/rewrite:
If a replacement string starts with “http://”, “https://”, or “$scheme”, the processing stops and the redirect is returned to a client.
If you want to just create the mapping between /desktop/$1 on the front-end and /apex/f?p=1001:$1 on the back-end of your Oracle Application Express (APEX), then the best way would be to use the following code on your nginx front-end server:
location /desktop/ {
rewrite ^/desktop/?(.*)$ /apex/f?p=1001:$1 break;
return 400;
proxy_pass http://smallblockpro.com:8080;
}
I would recommend copy-pasting it for each of /dev/, /mobile/ and /desktop/; also, I would not recommend to keep a slash-less versions, as per ServerFault's nginx-reverse-proxy-url-rewrite and how-to-remove-the-path-with-an-nginx-proxy-pass, as nginx already takes care of the requests without the trailing slash in a situation such as yours with the code as I propose above.
回答2:
Here's the copy-paste from what I'm using on our ORDS / SDW ( sqldev-web ) development server.
Here's a basic example with ORDS for the REST side of the house.
The access is to:
https://xyz.oraclecorp.com/sdw/klrice/metadata-catalog/
Then it's proxied to:
https://xyz.oraclecorp.com:8083/ords/klrice/metadata-catalog/
With this config. Beside not to rewrite to an absolute URI as that will do a full browser redirect vs just rewriting the url for the proxy pass.
location /sdw/ {
rewrite /sdw/(.*) /ords/$1 break;
proxy_pass https://xyz.oraclecorp.com:8083/ords/;
proxy_redirect off;
proxy_set_header Host $http_host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
}
The issue you will face is this
rewrite ^/desktop/(.*) http://smallblockpro.com:8080/apex/f?p=1001:$1 last;
APEX will see and write links/redirects/includes ( javascript/css/...) as the .../apex/XYZ which will hit the nginx server and that will not know what to do with a /apex/
Here's an example of that based on my above setup. Notice my request to /sdw/ turns into a Location redirect to /ords/
wget -S https://xyz.oraclecorp.com/sdw/
--2018-06-21 17:10:28-- https://xyz.oraclecorp.com/sdw/
Resolving xyz.oraclecorp.com... 123.456.789.123
Connecting to xyz.oraclecorp.com|123.456.789.123|:443... connected.
HTTP request sent, awaiting response...
HTTP/1.1 302 Found
Server: nginx/1.12.1
Location: https://xyz.oraclecorp.com/ords/f?p=4550:1:375440000433:::::
Location: https://xyz.oraclecorp.com/ords/f?p=4550:1:375440000433::::: [following]
So the easiest thing to do is match up the ords deployment ( /apex/ ) to what the rewrite/redirects are and use proxy pass to internalize the :8080 stuff. So
location ~ /desktop/ {
rewrite ^/desktop/(.*) http://smallblockpro.com/apex/f?p=1001:$1 last;
}
location ~ /apex/ {
proxy_pass http://smallblockpro.com:8080/apex/;
proxy_redirect off;
proxy_set_header Host $http_host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
}
This option will let your users have a nice entry point of /desktop/ but then that redirect the /apex/ for the app itself.
There is another option with ORDS url-mappings.xml to keep the /desktop/ also which would be something like this to add in mappings to ords so it knows the /desktop/. Then the nginx could do the same proxy pass for each of the entry urls.
url-mapping.xml file contents
<pool-config xmlns="http://xmlns.oracle.com/apex/pool-config">
<pool name="mypool" base-path="/desktop" />
</pool-config>
then in nginx
location ~ /desktop/ {
proxy_pass http://smallblockpro.com:8080/desktop/;
proxy_redirect off;
proxy_set_header Host $http_host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
}
回答3:
Before you read further go through below SO thread which explains about the extra /apex/
Nginx proxy_pass only works partially
Two issues in your config
- You need to pass the correct URL to backend service
- You need to make sure you handle any redirects and replace the url correctly
Below is the config I think should work for you
server {
listen 80 default_server;
listen [::]:80 default_server;
server_name example.com;
root /usr/share/nginx/html;
# Load configuration files for the default server block.
include /etc/nginx/default.d/*.conf;
location /dev/ {
proxy_pass http://example.com:8080/apex/;
proxy_redirect http://example.com:8080/apex/ $scheme://$host/dev/;
}
error_page 404 /404.html;
location = /40x.html {
}
error_page 500 502 503 504 /50x.html;
location = /50x.html {
}
}
来源：https://stackoverflow.com/questions/50848233/proxy-pass-ignoring-port