Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Browse files

Convert readme to markdown

  • Loading branch information...
commit e50a9eb29ed5c8eec71422b2b776d91742ce1d54 1 parent dde4504
@toddlipcon authored
Showing with 14 additions and 7 deletions.
  1. +14 −7 README →
@@ -1,6 +1,7 @@
Gremlins is a python program/framework which assists in fault testing distributed systems.
-== Overview ==
Gremlins is a fault injector - it is an evil program that sits on a machine and does
nasty things to other programs on that machine (or the machine overall).
@@ -14,7 +15,8 @@ The hope is that, if a fault tolerant system can make progress without corruptio
unrecoverable errors on a 5-node cluster full of gremlins, it will also be free of such
errors on a 500-node cluster where errors may occur every day by natural causes.
-== Setup ==
The very simplest way to run gremlins is to just set your python path:
@@ -25,11 +27,13 @@ after running python develop
$ gremlins
-== Concepts ==
Gremlins is built around a few key concepts: faults, triggers, and profiles.
-=== Faults ===
A fault is the most specific unit of abstraction - it is simply a thing that can go wrong.
One example fault is "kill -9 the DataNode JVM, wait 60 seconds, then start it back up".
@@ -46,7 +50,8 @@ other faults. Currently the only example of a metafault is gremlins.metafaults.p
which takes a list of (weight, fault) pairs, and picks one of the subfaults according to the
provided weights.
-=== Triggers ===
A trigger is a way of running faults. The simplest trigger (and the only one that really works
well at the moment) is gremlins.triggers.Periodic. This trigger is constructed with an interval
@@ -60,14 +65,16 @@ from a central location.
Future trigger ideas include the ability to watch a log for a given line before triggering a fault,
-=== Profiles ===
A profile currently doesn't have a type, but is just a normal python list of triggers. When
running gremlins, one can provide a profile, and each of the triggers will be started.
This concept allows one to start both a webserver trigger and a periodic trigger from a
single invocation.
-== Running gremlins ==
+Running gremlins
Gremlins can be run in two different modes. In the first mode, the user specifies a list of faults.
These faults are executed immediately, and then gremlins exists. In the second mode, the user
Please sign in to comment.
Something went wrong with that request. Please try again.