Experience of puppet usage for managment of Tier-1 grid cluster at NRC "Kurchatov Institute"

This article is about the organization of the cluster management using puppet. It tells about: safety of usage, from the point of view of mass apply at a computing cluster wrong configuration (by reason of human factor); collaboration work and the creation of opportunities for each cluster administrator, regardless of others, writing and debugging your own scripts, before include them in the overall system of cluster managment; writing scripts, which allow to get as fully configured nodes, and updates the configuration of any system parts, without affecting the rest of the nodes components, regardless of the current state of the node of computing cluster.
The article compares different methods of the creation of the hierarchy of puppet scenarios, describes problems associated with the use of “include” for the organization hierarchy, and tells about the transition to a system of sequential call classes through shell-script.