warewulf / wwmon Goto Github PK

Readme file for the Warewulf 3.x series

Earlier versions of Warewulf were focused on the entire cluster
solution. While this is the proper solution for many midsized HPC
clusters, it was not the right solution for many power users and other
provisioning requirements. It also limited its development.

Warewulf 3 implements a new paradigym. It seperates out the provisioning
subsystems from the user facilitating and management tools. Warewulf3 is
now strictly the distributed monitoring system. Provisoining is now
handled by a new solution called 'Perceus'.

Motivations for seperating the user tools from the provisioning
subsystem is as follows:

   - Some clusters may not implement a seperate monitoring from the
     schedule system. The provisioning model in Warewulf <= 2 required
     knowing exactly the status of the nodes because it managed an
     active role in management and syncronization.
   - We have had *many* requests to have the management subsystem
     seperated so it can be used to run on other cluster systems, other
     infrastuctures, etc...
   - Some people prefered the use of other cluster monitoring solutions.
     While we personally are biased to Warewulf's implementation, we
     understand that 1 size doesn't fit all.
   - This model ensures the greatest amount of flexiability and
     simplifies the management of the projects.

Warewulf is now strictly the user space and management cluster toolkit.
This toolkit is no longer dependant on how the cluster is maintained, or
which provisioning model is used. There are already people using the
Warewulf toolkit on non-Warewulf clusters, so this is simply making it
easier for them.

All node provisioning (eg. network booting, OS installation, etc) has
been migrated to a new project called Perceus. Perceus is a completly
new design offering a very advanced and platform neutral provisioning
mechanism. The model can be used to boot statefull or stateless nodes as
well as drive other mechanisms. It can also provision stateless clusters
and systems just as easy as an entire enterprise including workstations.
Its goals and design stay true to the original goals of the Warewulf
project, being: simple, intuitive, reliable, scalable, and as neutral to
the provisioned operating system as possiable. Perceus can be found at
http://www.perceus.org/.

The combination of Perceus and Warewulf allow for greater scalibility,
flexiability, and ease of use.