Building Highly Available LTSP Clusters

Revision 1.3: [ PDF | HTML | TXT ]
Source: [ OpenOffice ]

Abstract

Though this document as been created as a final grade project's requisite, and its release to the public should be considered as an acknowledgment and thanks to all the open-source community in general, and to be more precise, to all the Linux Terminal Server Project and Linux High Availability community, whose effort to achieve a stable status quo has been recognized during the project's implementation.

The goal of this document is to describe how to setup two LTSP servers to provide a Highly Available, Load Balanced LTSP service. This will be accomplished using the load-balancing and failover capabilities of the newer version of the ISC DHCPD daemon, and a storage replication mechanism for users profile availability on each server.

This approach will be phased in three steps: installing a working LTSP environment, and when everything is correctly set up, bring the cluster up and running.

Status

# Nodes	Load Balancer	Shared FS	Status	Notes
2	ISC DHCPD	DRBD	OK	LTSP 4.0, DRBD 0.6.12, Heartbeat 1.2.1
N	LVS for LTSP	*	TODO	All services would be load-balanced.
N	LVS for XDMCP	*	TODO	LVS would be balancing XDMCP -broadcast connections

(*) Note: Shared FS accomplished with distributed filesystems (DFS) or shared storage devices. Some DFS: CODA, (open)AFS, Intermezzo.