Project

General

Profile

Actions

User story #6278

open

Feature Idea: Fancy things with colors and state data, i.e. Dependency graphs

Added by Florian Heigl about 9 years ago. Updated almost 8 years ago.

Status:
New
Priority:
2
Assignee:
-
Category:
Web - Config management
UX impact:
Suggestion strength:
User visibility:
Effort required:
Name check:
Fix check:
Regression:

Description

Rudder has access & knowledge of a lot of information that could be "exploited" in the most wonderful ways.

Some examples
- Dependency graphs of NCF directives (i mean, look at class conditions!)
- Visual representation of datacenter-wide rule status

ginas had this with graphviz
https://github.com/ginas/ginas/blob/master/contrib/graphviz/ginas-role-dependencies.dot
and as an image
http://i.imgur.com/DeMRCSp.png

Actions #1

Updated by Florian Heigl about 9 years ago

  • Description updated (diff)

Forgot one key point from my scratch note, i'll just quote it

Export AND make accessible dependency graphs for node policies

Actions #2

Updated by Florian Heigl about 9 years ago

and the final twist...

it should be possible (first just in the academic meaning, but I think it really works) to use the combined info (class events + dependencies) with some fancy color thing like graphite to show a timelapse of changes.
the graphite idea is relevant to compare i.e. a spike in 500 errors and cfengine-triggered apache restarts following on the server issue.

summary of the thought:

Interface with analytics.

Actions #3

Updated by François ARMAND about 9 years ago

  • Priority changed from N/A to 2

I love that.

And so, we could say that Rudder is doing Big Data and Data mining, so big plus in buzz words (devops, cloud, big data is a full row in my bingo card).
More seriously, we need these dependency graph to be able to make more interesting stuff in techniques (or even generic method) and of course rule dependency management.

This will come along nicelly with the ability to tag things (rules, directives, groups, etc), and have analysis or reporting for them. Or even event-triggering based on the evolution of the analysis ("look, 42 % of my things tag "datacenter-XYZ have been in error in the last 6 hours, something is not going well")

I'm not sure exactly what we can provide and how to show it, but your propisition are a good starting point.

OK, that won't go realistically in 3.1, but I open in my text book a new major topic "do better analysis"

Actions #4

Updated by Benoît PECCATTE about 9 years ago

  • Category changed from 14 to Web - Config management
Actions #5

Updated by Benoît PECCATTE almost 8 years ago

  • Target version set to Ideas (not version specific)
Actions

Also available in: Atom PDF