“Less magic, more visibility” is what Puppet Labs reckons devops devotees need.
Announcing the company's Puppet Enterprise 2015.2, the company's director of marketing communications Tim Zonca told The Register the rapid turnaround ideal of the devops world has a trust problem that the new code is trying to address.
Someone running up VM instances based on pre-made code (think Docker) will take code from Puppet's forge, a GitHub repository or code shared by other organisations – but it's hard to see behind the magic, as Zonca put it.
That's the basis of the system's “Interactive Graph” function.
The idea is to give users a visual representation of their “infrastructure-as-code” models. Zonca explained to El Reg that “it's not uncommon for a single node – like a server or a virtual machine – to have a thousand or more “resources” on that node”.
Trawling through the code to understand its interactions with other nodes is a painful business, but so is spinning up a new node to find that it somehow upsets the surrounding environment.
“If you want to change some aspect of the configuration and it fails – what you didn't have before was a sense of, 'where does that live, and how does it relate to the other classes around it?'”
With the interactive graph, he said, “it's easier for the developer or the team to understand the context of the impact of part of the model.”
The interactive graph is part of a heavily reworked UI, Zonca said.
He also highlighted that the interactive graph reaches into the more abstract aspects of a deployment. He gave a WordPress instance by way of example: as well as WordPress, it will have a MySQL server, tables in the database, permissions granted to the database, plug-ins and more – each of which will be represented in the graph.
He called it an “emerging” graph, meaning that it continues interrogating models in Puppet Enterprise so it can keep up with changes and “show you the relationships that are in place.”
Also in the new release is an inventory filtering feature so users can search nodes and identify infrastructure of particular characteristics.
Users could, for example, capture features like a product's EOL, making it easer (for example) to identify Windows Server 2003 instances that need to be upgraded to WS 2012. The filter could also be used to capture and identify warranty information on each node.
The other major change Zonca pointed to is the unification of Puppet Enterprise's open-source and paid-for agents.
This gets rid of a wholesale change of agents deployed to nodes, if a user of the open source version decides to upgrade to the paid version.
Zonca said: “If you have 10,000 nodes running the open source agent, that's now the same agent. If you get a new master, all you have to do is point it at the nodes, because it's the same code.”
The company also reckons the new agent code is faster and occupies only half the memory.
The Puppet programming language has been simplified and enhanced with iteration, parameter type checking, and better error handling.
There's also new Puppet Supported modules, including a Cisco NX-OS agent and module for Nexus 3000 and 9000 switches (5000 and 7000 to follow); a module for Citrix NetScaler so Puppet Enterprise can support load balancing and content switching; and a module for VMware vSphere. ®