dice fault injection tool - fortissdownload.fortiss.org › ... › qudos2016 › slides ›...
TRANSCRIPT
DICE
Horizon 2020 Research & Innovation Action
Grant Agreement no. 644869
http://www.dice-h2020.eu
Funded by the Horizon 2020Framework Programme of the European Union
DICE Fault Injection Tool
Craig Sheridan, Darren Whigham
Flexiant Limited
Matej Artač
XLAB d.o.o.
Overview
o Motivation
o Fault Injection Tool
o DICE project and DevOps
o Conclusion
2©DICE
Motivation
o Infrastructure equipment fails
o Expensive approach:
Making hardware completely reliable
Making sure that software never fails
o Reasonable approach:
Make systems resilient to failures
o Need for on-demand failures
3©DICE
Fault Injection Tool4
©DICE
Fault Injection Tool
o Problem
When testing, it is difficult to reproduce faultshappening in production
o Goal
Causes environmental problems or failures on demand
Lets users control the levels and seriousness of induced failures
5/©DICE
Design
6©DICE
Supported VM-level Faults
• Shutdown random VM (black list or white list)
• High CPU for VM
• High Memory usage for VM
• Block VM external network access
• High Network Bandwidth usage
• High Disk I/O usage
• Stop a service running on VM
7©DICE
Supported Cloud-level Faults
o Shutdown a Node
o High CPU for Node
o High Memory usage for Node
o High Bandwidth usage
8/©DICE
Usage example command9
©DICE
$ fleximonkey --stressmem \2 \ test loops2048m \ total [email protected] \ ssh host-no \ no passwordhome/ubuntu/SSHKEYS/VMkey.key
Usage example result10
©DICE
Innovation
o ChaosMonkey: only accesses AWS & terminates VMs.
o Cocoma: developed but not supported, suffers from high resource usage, complex config & limited extensibility.
o FIT generates various cloud agnostic faults at VM & Cloud.
o Range of functions, greater flexibility to generate multiple faults
o Lightweight and only installs required tools and components on target VMs.
o Future extensibility in mind considering the needs and challenges of cloud service providers such as scalability and resiliency of the cloud consumption marketplace.
11©DICE
Big Data Technologies
Cloud (Priv/Pub)`
DICE Framework
12©DICE
DICE IDE
Profile
Plugins
Sim Ver Opt
DPIM
DTSM
DDSM TOSCAMethodology
Deploy Config Test
Mon
Anomaly
Trace
Iter. Enh.
Data Intensive Application (DIA)
Cont.Int. Fault Inj.
WP4
WP3
WP2
WP5
WP1 WP6 - Demonstrators
Use in DevOps
o Continuous Integration/Testing
Regular resilience tests
o Anomaly detection
Induce faults, check for anomalies
Help with quality enhancement
13/©DICE
Conclusion and future work
o Aimed at developers using quality driven DevOps approach
o Helps cloud operators perform stress analysis
o Difficulties and possibilities of extensibility by external users and investigating limitations
o Consider different topologies, operating systems and vendor agnostic cloud provider infrastructure
o Evaluate the overhead of operation
o Containerised environments will be a future target to help understand effect on microservices when injecting faults to the underlying host
14©DICE
Release: February 2017
https://github.com/dice-project/DICE-Fault-Injection-Tool
15/©DICE