httpd Posts

Run Multiple Containerized Keycloak Instances Behind an Apache HTTPD Proxy

On this session im trying to create a sample condition where im having a High-Availability cluster of multiple Keycloaks instances which are located behind an HTTPD Reverse Proxy. HTTPD will do a round robin request to two Keycloaks instances behind it, showing the capability of session sharing between different Keycloak instances.

Im using a containerized Keycloak image and run them by using Docker for simulating a condition of running more than one Keycloak instances with different IP.

The concept is pretty much like below image,

The first thing needed is setuping a database for this. In here im using MySQL, despite Keycloak is able to connect to different type of databases. And for this sample, database is installed on my host laptop and not using a containerized one.

CREATE USER 'keycloak'@'%' IDENTIFIED BY 'password';
CREATE DATABASE keycloak_db;

The next step is running two Keycloak instances by executing below command, in here im putting 192.168.56.1 as the ip for my host machine.

docker run -p 8081:8080 -e PROXY_ADDRESS_FORWARDING=true  \ 
	-e DB_VENDOR="mysql" -e DB_ADDR="192.168.56.1" -e DB_USER="keycloak" \ 
	-e DB_PASSWORD="password" -e DB_PORT="3306" -e DB_DATABASE="keycloak_db" \ 
	--add-host=HOST:192.168.56.1 jboss/keycloak


docker run -p 8082:8080 -e PROXY_ADDRESS_FORWARDING=true  \ 
	-e DB_VENDOR="mysql" -e DB_ADDR="192.168.56.1" -e DB_USER="keycloak" \ 
	-e DB_PASSWORD="password" -e DB_PORT="3306" -e DB_DATABASE="keycloak_db" \ 
	--add-host=HOST:192.168.56.1 jboss/keycloak

The good thing about this keycloak image is that by default it is running a standalone-ha.xml and automatically form a cluster when being run locally at the same time. This can be seen on Keycloak’s log

06:20:53,102 INFO  [org.infinispan.CLUSTER] (non-blocking-thread--p8-t4) 
	[Context=offlineClientSessions] ISPN100010: Finished rebalance with members [a629f48aafa9, 82298394e158], topology id 11
06:20:53,102 INFO  [org.infinispan.CLUSTER] (thread-35,ejb,a629f48aafa9) 
	[Context=sessions] ISPN100010: Finished rebalance with members [a629f48aafa9, 82298394e158], topology id 11
06:20:53,103 INFO  [org.infinispan.CLUSTER] (thread-30,ejb,a629f48aafa9) 
	[Context=work] ISPN100010: Finished rebalance with members [a629f48aafa9, 82298394e158], topology id 11
06:20:53,114 INFO  [org.infinispan.CLUSTER] (thread-30,ejb,a629f48aafa9) 
	[Context=loginFailures] ISPN100010: Finished rebalance with members [a629f48aafa9, 82298394e158], topology id 11
06:20:53,121 INFO  [org.infinispan.CLUSTER] (thread-31,ejb,a629f48aafa9) 
	[Context=actionTokens] ISPN100010: Finished rebalance with members [a629f48aafa9, 82298394e158], topology id 11

The last step would be creating an HTTPD setup for creating reverse proxy with a load-balancing capability. And we can achieve that by editing httpd.conf file, for this sample we are using a round-robin mechanism of lbmethod=byrequests.

<VirtualHost *:80>
	ServerName localhost
	ProxyRequests Off
	ProxyPreserveHost On
  
	<Proxy "balancer://mycluster">
		BalancerMember http://localhost:8081
		BalancerMember http://localhost:8082
	
		ProxySet lbmethod=byrequests failontimeout=on
	 
	</Proxy>
  
	ProxyPass / balancer://mycluster/
	ProxyPassReverse / balancer://mycluster/
</VirtualHost>

In order to activate loadbalancer and http proxying feature on Apache HTTPD, there are several variables that need to be unremark on httpd.conf file such as proxy_balancer_module and proxy_http_module.

Restart httpd and open browser directly, and we can see that Keycloak is working well.

We can also simulate a condition where one instance suddenly stopped to simulate a failover scenario by killing one Keycloak instance using docker kill command. Have fun with Keycloak 😀

Notes.
For a better performance, a sticky session approach is recommended compared to a round robin one.

https://www.keycloak.org/docs/latest/server_installation/#sticky-sessions

Error Apache HTTPD, No protocol handler was valid for the URL

Got weird error today when implementing apache httpd as a reverse proxy on my RHEL environment, here is the complere stacktrace

AH01144: No protocol handler was valid for the URL /url (scheme 'http'). 
If you are using a DSO version of mod_proxy, make sure the proxy submodules are included in the configuration using LoadModule.

The only module that i unremark on httpd.conf is mod_proxy.so

LoadModule proxy_module modules/mod_proxy.so

I found out that i need to unremark other module, that is mod_proxy_http

LoadModule proxy_http_module modules/mod_proxy_http.so

After unremarking mod_proxy_http, everything runs well.