Umbrella repository for the various DepTrack projects.
DepTrack is an approach more than a tool. The key finding of the DepTrack approach is that programs logs are more useful when shaped like a tree than when shaped like a sequence. Using such an approach, rather than manually connecting nodes in a graph, one can focus on computing a value which describes a high-level goal and records a tree of dependencies used during the computation.
The lowest-level brick, a monad transformer to describe trees and directed-acyclic graphs. This brick allows you to write DSL-like libraries that record directed-acyclic graphs with a node type that best describe your problem. This approach let you decouple how you operate on (i.e., what you do with) the graph from express (i.e., which library builds) the graph. This way, the day you feel like DepTrack is no longer suitable to express your dependency graph you can still keep your functions to operate on the graph.
This library uses deptrack-core to perform Devops tasks. Devops is a job title, and Devops engineers need to do two main things: (a) describe an infrastructure and, (b) evaluate the description (e.g., to make the world converge towards their intention, or to diagnose which bits of the system diverged too much). In particular, (a) encodes a dependency graph between various subsystems of the infrastructure: faults and outages follow dependencies. Hence, the Devops space is a good playground for the DepTrack approach of building dependency graphs.
The author was dissatisfied with existing Devops tools: some existing tools claim to use a declarative syntax (generally YAML) but actually use templated-YAML. As a result, the actual description of an infrastructure (a) is a mix of a template language and a description language. The language used in the evaluator (b) often is the same scripting language as the one used in the description, however they do not interact well: a main program provided by the solution is in charge of generating the description and evaluating it. Such main programs impose structure (which can be useful at times) but leave little room for creativity. Last, the fact that scripting languages dominate the Devops tools space is annoying because applications written with scripting languages are ironically hard to deploy and have poor performance (and abismal multicore scalability).
Existing tools have a number of drawbacks:
Our industry needs to empower Devops to describe infrastructure and operate on the description creatively (e.g., generating statistics/simulations from such descriptions). DepTrack is an attempt to move the state of the art forward. Once you have a graph and manipulate it, you can estimate resource needs, you can budget your monitoring on the most critical bits etc.
This library is written in Haskell and requires Devops to also describe infrastructure in Haskell. Using Haskell for DepTrack has the following advantages:
One drawback of using Haskell is that there's more upfront learning than learning YAML+scripting languages. However, longer term, the author believes you can reach further with DepTrack than existing Devops tools.
This module provides some recipes to configure machines. It contains things I found useful to configure my personal servers; mostly Debian-like. Organization of this module is likely to change. See the README in this directory to have an up-to date description.
We provide a Mustache template to start new DepTrack for Devops projects. After installing the Haskell Stack tool, you can do as follows:
stack new my-project https://raw.githubusercontent.com/lucasdicioccio/deptrack-project/master/deptrack-devops.hsfiles cd my-project stack build
You can also study and modify the examples from deptrack-devops-examples.
This can be done as follows
git clone https://github.com/lucasdicioccio/deptrack-project.git cd deptrack-project stack setup cd deptrack-devops-examples stack install
Contributions to DepTrack projects are welcome. Please use GitHub issues for discussions. Please add yourself to the Contributors list in your pull requests.
The bulleted-list of the DepTrack contributor is as follows:
Sorted by first name: