Thread: HA distributed linux cluster

We're currently investigating splitting our co-lo kit across a number of sites to reduce the impact of outages. heartbeat / DRBD looks like an ideal solution, just wondered if anyone had any experience or pointers before I start. How does it scale? Is Piranha worth a look? Anything else out there (GNU/GPL!)

Been testing heartbeat / DRBD for a few applications for the last few months, workings flawless so far!

A downside to DRBD is the fact you can't do anything with the DRBD device if the server is in secondary mode, but that's not a problem for me. It can't be used for load balancing, but its very good for failover!

DRBD is nice, but as Luke says, you can't write to the secondary server.

That might not be too much of a problem for some applications (static web server, or web server where data is written to a seperate database server) though.

The standard solution I provide for people uses LVS - two front-end controllers running heartbeat for automated failover, and 3 or more back-end servers. This has the added benefit of giving you scalability as well as just protection against outages. For back-end storage I mount via NFS, with the storage server either being a Linux-server with raid array or a NetApp for more heavy duty usage.

The best solution for you rather depends on exactly what it is you're trying to replicate of course.