LVSM Introduction

LVSM is a collection of software packages that turn a collection of linux machines into a scalable, highly available server farm.

LVSM is built on a number of free software projects:

LVS

LVSM uses the ipvs kernel patches and the ipvsadm utility from the lvs project to perform the level four switching.

Heartbeat

LVSM uses heartbeat for monitoring node health, providing vip failover, and for reliable cluster communications.

MON

LVSM uses mon to monitor service health. MON is very general, so supporting a new (or custom) application is as easy as writing a script that can tell if the service in question is functioning.

In addition, LVSM contains some new software:

PLVSD

The Penguin LVS Daemon. This is the piece that glues the whole thing together. Plvsd is in charge of maintaining the kernel routing tables for the virtual services supported, as well as providing for remote adminstration and on the fly configuration.

Web GUI

LVSM contains a web based (mod_perl) gui which can talk to plvsd on any machine in the cluster. From this gui you can view cluster resources and health, change the cluster configuration on the fly, as well as see realtime statistics about the cluster.

Depending upon which role each machine serves, it needs to have a different subset of these packages installed. All machines in the cluster need to be running heartbeat and plvsd. An LVSM installation consists of the following node types:

Director

Director nodes are those which schedule and forward incoming requests to realservers. As such, they need to be running the appropriate ipvs kernel patches. Also, the director node also runs mon to keep track of service status on the realservers.

Realserver

Realserver nodes are those which host the actual service being run on the cluster.

Admin Webserver

In addition, there needs to be a machine on which the web gui runs. This machine may or may not be part of the cluster. It needs only to have tcp connectivity to any of the cluster machines.