Apache Clustering

Continuous uptime is critical to todays ebusiness applications. LifeKeeper monitors Apache and all dependent resources down to hardware component level. This enables failover of Apache web servers and recovery of websites instantly.

Failover can be across a LAN with shared storage or, without shared storage if data replication is used.

LifeKeeper reduces the costs associated with clustering in two ways. Firstly, by eliminating the requirement for shared storage, and secondly by eliminating the requirement for a redundant server. The backup server is run as an active server and can run another application.

Protecting Apache without shared storage

In a typical local configuration, nothing is shared between the servers. Identical copies of the Apache Web Server configuration file, web documents, DSO modules (and their configuration files, if any), and the httpd executable reside in exactly the same locations on each server.

Each web site is assigned an IP address – or a domain address that maps to a particular IP address - through the configuration file, and a LifeKeeper IP address is created for each IP address. When the Apache hierarchy is switched over from one server to another, this particular httpd instance is stopped and the IP addresses are deactivated on the first server, then the IP addresses are reactivated and the instance started on the other server. Clients will then be automatically connected via TCP/IP to the identical web site on the other server.

Recovering Apache servers using Shared file systems

Apache web servers can be recovered using shared storage (below). The example below shows an active/active environment where the Apache server is acting as a backup for a SAMBA server. This eliminates the need to have a redundant server as a backup server.

Failover in an n+1 environment

Multiple servers can be protected simultaneously by setting up an N+1 configuration. The example below shows a typical server farm running multiple apache servers. These servers can be protected by a single backup server that will monitor all of the servers in the cluster. LifeKeeper allows up to 32 nodes in a cluster.

Disaster recovery to a remote site

To protect against a site disaster, failover to another site is achieved by recovering to server located in a separate location.

The scenario below shows data replication taking place across a WAN to a backup server in a remote location. When Apache fails over, the data is up to date.

Original Source: www.openminds.co.uk