Posts Tagged ‘incoming requests’

How to Configure Nginx as a Reverse Proxy Load Balancer on Debian, CentOS, RHEL Linux

Monday, December 14th, 2020

Reading Time: 8minutes

set-up-nginx-reverse-proxy-howto-linux-logo

What is reverse Proxy?

Reverse Proxy (RP) is a Proxy server which routes all incoming traffic to secondary Webserver situated behind the Reverse Proxy site.

Then all incoming replies from secondary webserver (which is not visible) from the internet gets routed back to Reverse Proxy service. The result is it seems like all incoming and outgoing HTTP requests are served from Reverse Proxy host where in reality, reverse proxy host just does traffic redirection. Problem with reverse proxies is it is one more point of failure the good side of it can protect and route only certain traffic to your webserver, preventing the behind reverse proxy located server from crackers malicious HTTP requests.

Treverse proxy, which accepts all traffic and forwards it to a specific resource, like a server or container.  Earlier I've blogged on how to create Apache reverse Proxy to Tomcat.
Having a reverse proxy with apache is a classical scenarios however as NGINX is taking lead slowly and overthrowing apache's use due to its easiness to configure, its high reliability and less consumption of resources.


One of most common use of Reverse Proxy is to use it as a software Load Balancer for traffic towards another webserver or directly to a backend server. Using RP as a to mitigate DDoS attacks from hacked computers Bot nets (coming from a network range) is very common Anti-DDoS protection measure.
With the bloom of VM and Contrainerizations technology such as docker. And the trend to switch services towards micro-services, often reverse proxy is used to seamessly redirect incoming requests traff to multiple separate OS docker running containers etc.


Some of the other security pros of using a Reverse proxy that could be pointed are:

  • Restrict access to locations that may be obvious targets for brute-force attacks, reducing the effectiveness of DDOS attacks by limiting the number of connections and the download rate per IP address. 
  • Cache pre-rendered versions of popular pages to speed up page load times.
  • Interfere with other unwanted traffic when needed.

 


what-is-reverse-proxy-explained-proxying-tubes

 

1. Install nginx webserver


Assuming you have a Debian at hand as an OS which will be used for Nginx RP host, lets install nginx.
 

[hipo@proxy ~]$ sudo su – root

[root@proxy ~]#  apt update

[root@proxy ~]# apt install -y nginx


Fedora / CentOS / Redhat Linux could use yum or dnf to install NGINX
 

[root@proxy ~]# dnf install -y nginx
[root@proxy ~]# yum install -y nginx

 

2. Launch nginx for a first time and test


Start nginx one time to test default configuration is reachable
 

systemctl enable –now nginx


To test nginx works out of the box right after install, open a browser and go to http://localhost if you have X or use text based browser such as lynx or some web console fetcher as curl to verify that the web server is running as expected.

nginx-test-default-page-centos-linux-screenshot
 

3. Create Reverse proxy configuration file

Remove default Nginx package provided configuration

As of 2020 by default nginx does load configuration from file /etc/nginx/sites-enabled/default on DEB based Linuxes and in /etc/nginx/nginx.conf on RPM based ones, therefore to setup our desired configuration and disable default domain config to be loaded we have to unlink on Debian

[root@proxy ~]# unlink /etc/nginx/sites-enabled/default

or move out the original nginx.conf on Fedora / CentOS / RHEL:
 

[root@proxy ~]# mv /etc/nginx/nginx.conf /etc/nginx/nginx.conf-distro

 

Lets take a scenario where you have a local IP address that's NAT-ted ot DMZ-ed and and you want to run nginx to server as a reverse proxy to accelerate traffic and forward all traffic to another webserver such as LigHTTPD / Apache or towards java serving Application server Jboss / Tomcat etc that listens on the same host on lets say on port 8000 accessible via app server from /application/.

To do so prepare new /etc/nginx/nginx.conf to look like so
 

[root@proxy ~]# mv /etc/nginx/nginx.conf /etc/nginx.conf.bak
[root@proxy ~]# vim /etc/nginx/nginx.conf

user nginx;
worker_processes auto;
worker_rlimit_nofile 10240;
pid /run/nginx.pid;
include /etc/nginx/modules-enabled/*.conf;

 

events {
#       worker_connections 768;
        worker_connections 4096;
        multi_accept on;
        # multi_accept on;
}

http {

        ##
        # Basic Settings
        ##

        sendfile on;
        tcp_nopush on;
        tcp_nodelay on;
        #keepalive_timeout 65;
        keepalive_requests 1024;
        client_header_timeout 30;
        client_body_timeout 30;
        keepalive_timeout 30;
        types_hash_max_size 2048;
        # server_tokens off;

        # server_names_hash_bucket_size 64;
        # server_name_in_redirect off;

        include /etc/nginx/mime.types;
        default_type application/octet-stream;

        ##
        # SSL Settings
        ##

        ssl_protocols TLSv1 TLSv1.1 TLSv1.2; # Dropping SSLv3, ref: POODLE
        ssl_prefer_server_ciphers on;

        ##
        # Logging Settings
        ##

        access_log /var/log/nginx/domain.com/access.log;
        error_log /var/log/nginx/domain.com/error.log;

        ##
        # Gzip Settings
        ##

        gzip on;

        # gzip_vary on;
        # gzip_proxied any;
        # gzip_comp_level 6;
        # gzip_buffers 16 8k;
        # gzip_http_version 1.1;
        # gzip_types text/plain text/css application/json application/javascript text/xml application/xml application/xml+rss text/javascript;

        ##
        # Virtual Host Configs
        ##

        include /etc/nginx/conf.d/*.conf;
        include /etc/nginx/sites-enabled/*;

 include /etc/nginx/default.d/*.conf;

upstream myapp {
    server 127.0.0.1:8000 weight=3;
    server 127.0.0.1:8001;
    server 127.0.0.1:8002;
    server 127.0.0.1:8003;
# Uncomment and use Load balancing with external FQDNs if needed
#  server srv1.example.com;
#   server srv2.example.com;
#   server srv3.example.co

}

#mail {
#       # See sample authentication script at:
#       # http://wiki.nginx.org/ImapAuthenticateWithApachePhpScript
#
#       # auth_http localhost/auth.php;
#       # pop3_capabilities "TOP" "USER";
#       # imap_capabilities "IMAP4rev1" "UIDPLUS";
#
#       server {
#               listen     localhost:110;
#               protocol   pop3;
#               proxy      on;
#       }
#
#       server {
#               listen     localhost:143;
#               protocol   imap;
#               proxy      on;
#       }
#}

 

In the example above, there are 3 instances of the same application running on 3 IPs on different ports, just for the sake to illustrate Fully Qualified Domain Names (FQDNs) Load balancing is also supported you can see the 3 commented application instances srv1-srv3.
 When the load balancing method is not specifically configured, it defaults to round-robin. All requests are proxied to the server group myapp1, and nginx applies HTTP load balancing to distribute the requests.Reverse proxy implementation in nginx includes load balancing for HTTP, HTTPS, FastCGI, uwsgi, SCGI, memcached, and gRPC.
To configure load balancing for HTTPS instead of HTTP, just use “https” as the protocol.


To download above nginx.conf configured for High traffic servers and supports Nginx virtualhosts click here.

Now lets prepare for the reverse proxy nginx configuration a separate file under /etc/nginx/default.d/ all files stored there with .conf extension are to be red by nginx.conf as instructed by /etc/nginx/nginx.conf :

We'll need prepare a sample nginx

[root@proxy ~]# vim /etc/nginx/sites-available/reverse-proxy.conf

server {

        listen 80;
        listen [::]:80;


 server_name domain.com www.domain.com;
#index       index.php;
# fallback for index.php usually this one is never used
root        /var/www/domain.com/public    ;
#location / {
#try_files $uri $uri/ /index.php?$query_string;
#}

        location / {
                    proxy_pass http://127.0.0.1:8080;
  }

 

location /application {
proxy_pass http://domain.com/application/ ;

proxy_http_version                 1.1;
proxy_cache_bypass                 $http_upgrade;

# Proxy headers
proxy_set_header Upgrade           $http_upgrade;
proxy_set_header Connection        "upgrade";
proxy_set_header Host              $host;
proxy_set_header X-Real-IP         $remote_addr;
proxy_set_header X-Forwarded-For   $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_set_header X-Forwarded-Host  $host;
proxy_set_header X-Forwarded-Port  $server_port;

# Proxy timeouts
proxy_connect_timeout              60s;
proxy_send_timeout                 60s;
proxy_read_timeout                 60s;

        access_log /var/log/nginx/reverse-access.log;
        error_log /var/log/nginx/reverse-error.log;

}

##listen 443 ssl;
##    ssl_certificate /etc/letsencrypt/live/domain.com/fullchain.pem;
##    ssl_certificate_key /etc/letsencrypt/live/domain.com/privkey.pem;
##    include /etc/letsencrypt/options-ssl-nginx.conf;
##    ssl_dhparam /etc/letsencrypt/ssl-dhparams.pem;

}

Get above reverse-proxy.conf from here

As you see config makes all incoming traffic towards root ( / ) NGINX directory for domain http://domain.com on port 80 on Nginx Webserver to be passed on http://127.0.0.1:8000/application.

      location / {
                    proxy_pass http://127.0.0.1:8000;
  }


Another set of configuration has configuration domain.com/application to reverse proxy to Webserver on Port 8080 /application.

 

location /application {
proxy_pass http://domain.com/application/ ;

proxy_http_version                 1.1;
proxy_cache_bypass                 $http_upgrade;

# Proxy headers
proxy_set_header Upgrade           $http_upgrade;
proxy_set_header Connection        "upgrade";
proxy_set_header Host              $host;
proxy_set_header X-Real-IP         $remote_addr;
proxy_set_header X-Forwarded-For   $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_set_header X-Forwarded-Host  $host;
proxy_set_header X-Forwarded-Port  $server_port;

# Proxy timeouts
proxy_connect_timeout              60s;
proxy_send_timeout                 60s;
proxy_read_timeout                 60s;

        access_log /var/log/nginx/reverse-access.log;
        error_log /var/log/nginx/reverse-error.log;

}

– Enable new configuration to be active in NGINX

 

[root@proxy ~]# ln -s /etc/nginx/sites-available/reverse-proxy.conf /etc/nginx/sites-enabled/reverse-proxy.conf

 

4. Test reverse proxy nginx config for syntax errors

 

[root@proxy ~]# nginx -t

 

nginx: the configuration file /etc/nginx/nginx.conf syntax is ok
nginx: configuration file /etc/nginx/nginx.conf test is successful

Test connectivity to listen external IP address

 

5. Enable nginx SSL letsencrypt certificates support

 

[root@proxy ~]# apt-get update
[root@proxy ~]# apt-get install software-properties-common

[root@proxy ~]# apt-get update
[root@proxy ~]# apt-get install python-certbot-nginx

 

6. Generate NGINX Letsencrypt certificates

 

[root@proxy ~]# certbot –nginx

Saving debug log to /var/log/letsencrypt/letsencrypt.log
Plugins selected: Authenticator nginx, Installer nginx

Which names would you like to activate HTTPS for?
– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –
1: your.domain.com
– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –
Select the appropriate numbers separated by commas and/or spaces, or leave input
blank to select all options shown (Enter 'c' to cancel): 1
Obtaining a new certificate
Performing the following challenges:
http-01 challenge for your.domain.com
Waiting for verification…
Cleaning up challenges
Deploying Certificate to VirtualHost /etc/nginx/sites-enabled/reverse-proxy.conf

Please choose whether or not to redirect HTTP traffic to HTTPS, removing HTTP access.
– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –
1: No redirect – Make no further changes to the webserver configuration.
2: Redirect – Make all requests redirect to secure HTTPS access. Choose this for
new sites, or if you're confident your site works on HTTPS. You can undo this
change by editing your web server's configuration.
– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –
Select the appropriate number [1-2] then [enter] (press 'c' to cancel): 2
Redirecting all traffic on port 80 to ssl in /etc/nginx/sites-enabled/reverse-proxy.conf

– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –
Congratulations! You have successfully enabled https://your.domain.com

You should test your configuration at:
https://www.ssllabs.com/ssltest/analyze.html?d=your.domain.com
– – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –

 

7. Set NGINX Reverse Proxy to auto-start on Linux server boot

On most modern Linux distros use systemctl for legacy machines depending on the Linux distribution use the addequate runlevel /etc/rc3.d/ symlink link on Debian based distros on Fedoras / CentOS / RHEL and other RPM based ones use chkconfig RedHat command.

 

[root@proxy ~]# systemctl start nginx
[root@proxy ~]# systemctl enable nginx

 

8. Fixing weird connection permission denied errors


If you get a weird permission denied errors right after you have configured the ProxyPass on Nginx and you're wondering what is causing it you have to know by default on CentOS 7 and RHEL you'll get this errors due to automatically enabled OS selinux security hardening.

If this is the case after you setup Nginx + HTTPD or whatever application server you will get errors in  /var/log/nginx.log like:

2020/12/14 07:46:01 [crit] 7626#0: *1 connect() to 127.0.0.1:8080 failed (13: Permission denied) while connecting to upstream, client: 127.0.0.1, server: , request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:8080/", host: "localhost"
2020/12/14 07:46:01 [crit] 7626#0: *1 connect() to 127.0.0.1:8080 failed (13: Permission denied) while connecting to upstream, client: 127.0.0.1, server: , request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:8080/", host: "localhost"
2020/12/14 07:46:01 [crit] 7626#0: *1 connect() to 127.0.0.1:8080 failed (13: Permission denied) while connecting to upstream, client: 127.0.0.1, server: , request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:8080/", host: "localhost"
2020/12/14 07:46:02 [crit] 7626#0: *1 connect() to 127.0.0.1:8080 failed (13: Permission denied) while connecting to upstream, client: 127.0.0.1, server: , request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:8080/", host: "localhost"


The solution to proxy_pass weird permission denied errors is to turn off selinux

[root@proxy ~]# setsebool -P httpd_can_network_connect 1

To permanently allow nginx and httpd

[root@proxy ~]# cat /var/log/audit/audit.log | grep nginx | grep denied | audit2allow -M mynginx
[root@proxy ~]# semodule -i mynginx.pp

 

[root@proxy ~]# cat /var/log/audit/audit.log | grep httpd | grep denied | audit2allow -M myhttpd
[root@proxy ~]# semodule -i myhttpd.pp


Then to let know nginx and httpd (or whatever else app you run) be aware of new settings restart both

[root@proxy ~]# systemctl restart nginx
[root@proxy ~]# systemctl restart httpd

Auto restart Apache on High server load (bash shell script) – Fixing Apache server temporal overload issues

Saturday, March 24th, 2012

Reading Time: 4minutes

auto-restart-apache-on-high-load-bash-shell-script-fixing-apache-temporal-overload-issues

I've written a tiny script to check and restart, Apache if the server encounters, extremely high load avarage like for instance more than (>25). Below is an example of a server reaching a very high load avarage:;

server~:# uptime
13:46:59 up 2 days, 18:54, 1 user, load average: 58.09, 59.08, 60.05
load average: 0.09, 0.08, 0.08

Sometimes high load avarage is not a problem, as the server might have a very powerful hardware. A high load numbers is not always an indicator for a serious problems. Some 16 CPU dual core (2.18 Ghz) machine with 16GB of ram could probably work normally with a high load avarage like in the example. Anyhow as most servers are not so powerful having such a high load avarage, makes the machine hardly do its job routine.

In my specific, case one of our Debian Linux servers is periodically reaching to a very high load level numbers. When this happens the Apache webserver is often incapable to serve its incoming requests and starts lagging for clients. The only work-around is to stop the Apache server for a couple of seconds (10 or 20 seconds) and then start it again once the load avarage has dropped to less than "3".

If this temporary fix is not applied on time, the server load gets increased exponentially until all the server services (ssh, ftp … whatever) stop responding normally to requests and the server completely hangs …

Often this server overloads, are occuring at night time so I'm not logged in on the server and one such unexpected overload makes the server unreachable for hours.
To get around the sudden high periodic load avarage server increase, I've written a tiny bash script to monitor, the server load avarage and initiate an Apache server stop and start with a few seconds delay in between.

#!/bin/sh
# script to check server for extremely high load and restart Apache if the condition is matched
check=`cat /proc/loadavg | sed 's/\./ /' | awk '{print $1}'`
# define max load avarage when script is triggered
max_load='25'
# log file
high_load_log='/var/log/apache_high_load_restart.log';
# location of inidex.php to overwrite with temporary message
index_php_loc='/home/site/www/index.php';
# location to Apache init script
apache_init='/etc/init.d/apache2';
#
site_maintenance_msg="Site Maintenance in progress - We will be back online in a minute";
if [ $check -gt "$max_load" ]; then>
#25 is load average on 5 minutes
cp -rpf $index_php_loc $index_php_loc.bak_ap
echo "$site_maintenance_msg" > $index_php_loc
sleep 15;
if [ $check -gt "$max_load" ]; then
$apache_init stop
sleep 5;
$apache_init restart
echo "$(date) : Apache Restart due to excessive load | $check |" >> $high_load_log;
cp -rpf $index_php_loc.bak_ap $index_php_loc
fi
fi

The idea of the script is partially based on a forum thread – Auto Restart Apache on High Loadhttp://www.webhostingtalk.com/showthread.php?t=971304Here is a link to my restart_apache_on_high_load.sh script

The script is written in a way that it makes two "if" condition check ups, to assure 100% there is a constant high load avarage and not just a temporal 5 seconds load avarage jump. Once the first if is matched, the script first tries to reduce the server load by overwritting a the index.php, index.html script of the website with a one stating the server is ongoing a maintenance operations.
Temporary stopping the index page, often reduces the load in 10 seconds of time, so the second if case is not necessery at all. Sometimes, however this first "if" condition cannot decrease enough the load and the server load continues to stay too high, then the script second if comes to play and makes apache to be completely stopped via Apache init script do 2 secs delay and launch the apache server again.

The script also logs about, the load avarage encountered, while the server was overloaded and Apache webserver was restarted, so later I can check what time the server overload occured.
To make the script periodically run, I've scheduled the script to launch every 5 minutes as a cron job with the following cron:

# restart Apache if load is higher than 25
*/5 * * * * /usr/sbin/restart_apache_on_high_load.sh >/dev/null 2>&1

I have also another system which is running FreeBSD 7_2, which is having the same overload server problems as with the Linux host.
Copying the auto restart apache on high load script on FreeBSD didn't work out of the box. So I rewrote a little chunk of the script to make it running on the FreeBSD host. Hence, if you would like to auto restart Apache or any other service on FreeBSD server get /usr/sbin/restart_apache_on_high_load_freebsd.sh my script and set it on cron on your BSD.

This script is just a temporary work around, however as its obvious that the frequency of the high overload will be rising with time and we will need to buy new server hardware to solve permanently the issues, anyways, until this happens the script does a great job 🙂

I'm aware there is also alternative way to auto restart Apache webserver on high server loads through using monit utility for monitoring services on a Unix system. However as I didn't wanted to bother to run extra services in the background I decided to rather use the up presented script.

Interesting info to know is Apache module mod_overload exists – which can be used for checking load average. Using this module once load avarage is over a certain number apache can stop in its preforked processes current serving request, I've never tested it myself so I don't know how usable it is. As of time of writting it is in early stage version 0.2.2
If someone, have tried it and is happy with it on a busy hosting servers, please share with me if it is stable enough?