Topic: Redundant Servers

[Jon Blakely]

I am currently producing a proposal to provide a global franchise operation with a SME server hosting their websites, secure email, webmail, mailists etc. One of the main requirements of the customer is a maximum of 1hr downtime.

The server will be housed in a colocation facility with redundant power (clean mains,UPS and Generator), redundant and diverse 155Mbit fibres from two telcos into the building and four downstream bandwidth suppliers.

Basically that leaves hardware as the major risk of a sustained outage. I plan to install two identical servers, primary and secondary, each with its own static ip address and use fail-over DNS to detect a failure of the primary and point to the secondary.

Here-in lies the problem. I need to reflect any changes made in the primary server to the secondary server. I can use Unison to synchronise all user data, ibays etc but I also need to synchronise  changes to users, groups etc made via server-manager.
Using Unison to synchronise the whole server i.e / less excludes like /dev, /proc /tmp and any other special files would synchronise all changes across but would also synchronise the configuration and account database which includes ip address and host name so this wouldn't work.

Anyone doing anything like this or have any suggestions.

Jon

[bobk]

There was a High Availability project being developed. I am not sure of the current status but you can take a look at http://www.star-support.com/sme/Linux-HA/SME%20High%20Availability%20How-To.html.

Contact the developer for details on status of the project and any up-dates.

[Jon Blakely]

Had looked at that and found it was not suitable. While it provides hardware redundancy via a heartbeat between the 2 machines it creates a network raid data disk between the two systems but does not synchronise system files.

Jon