Continuous Integration
with Hudson
Arun Gupta
Sun Microsystems, Inc.

Acknowledgements
Originally presented by
Kohsuke Kawaguchi & Jesse Glick
at JavaOne
Scheduled to be presented by
Fabiane Nardon
At FISL 10

Rise of Continuous Integration
> Offload from people, push to computers
$
computers
us
time
3

“Never use humans for the job of a machine”

Spend more CPU power to help you
> … even if it only helps a little
> First on your laptops and workstations
 IDEs are at the forefront
> And then to the servers
 a.k.a. “Continuous Integration”
 More frequent build/test executions
 Static code analysis tools
 And more to come
5

Hudson https://hudson.dev.java.
net/
> Open-source CI server at java.net
> Emphasis on ease of installation and use
 “java -jar hudson.war” execution
 Configure everything from browsers
> Extensibility
 140+ community-developed public plugins
 By 150+ contributors
> Estimated 13,000 installations
6

It basically does builds and tests
> Check out the source code
 Subversion, Perforce, Git, Mercurial, CVS, …
> Do builds and/or tests
 Ant, Maven, MSBuild, shell script, …
> Record results
 Binary, test results, code coverage, static analysis
> Notify people
 E-mail, IM, RSS, tray apps, IDEs
7

Hudson Plugin Ecosystem
New Plugins
2008: 55
2009: 44 so far
8

Localized to 8 languages
9

And hopefully more to come…
10

Adoption in all kinds of businesses
11

Eclipse Community Survey
12

13

Why Distributed Builds?
> You need to use multiple computers because…
 You need different environments
 You need isolation
> There’s only so much you can do with 1 computer
14

Installing new slaves
> For first 20 or so slaves, we did it manually
 Insert CD, click, type, click, type, click, …
 But that doesn’t scale
> Then we automated
 Available as “Hudson PXE Plugin”
15

Automated System Installations
> Hudson + PXE plugin
 ISO images of OS
> Slaves
 Power on, hit F12
 PC boots from network (PXE)
16

Automated System Installations
> Hudson + PXE plugin
 ISO images of OS
Your corporate IT guy &
his DHCP server
> Slaves
 Power on, hit F12
 PC boots from network (PXE)
 Choose OS from menu
 Installs non-interactively
17

Automated System Installations
> Supports OpenSolaris, Ubuntu, CentOS, Fedora
 Trivial with most Linux
> Cooperate with Windows, too
> Quite useful outside Hudson, too
 No more broken CD drives
 No more CD-Rs
18

Distributed builds with Hudson
> Master
 Serves HTTP requests
 Stores all important info
> Slaves
 170KB single JAR
 Assumed to be unreliable
 Scale to at least 100
> Link
 Single bi-di byte stream
 No other requirements
19

Heterogeneous Cluster Challenge
> Your builds/tests need to run in specific
environment
> Dependency on individual nodes hurts utilization
jobs slaves
Wombat Window
Windows test
s #1
GlassFish Window
Windows test
s #2
Hudson Solaris
Windows test
#1
Hudson Solaris
test
20

Labels to rescue
> Label is a group of slaves
> Tie jobs to labels
jobs slaves
Wombat Window
Windows test
s #1
Windows
GlassFish Window
Windows test
s #2
Hudson Solaris
Windows test
Solaris #1
Hudson Solaris Window
test
s #3
21

Coming to
your
Hudson
Making builds sticky soon
> We want jobs to be mostly on the same slave
 Faster check out
 Consistent results
 Minimizes disk consumption
> But does it softly
> Hudson uses consistent hash* to achieve this
> More schedule controls become possible:
 Use faster machines more frequently
 Slowly ramp up newly installed slaves
* http://en.wikipedia.org/wiki/Consistent_hashing 22

Forecasting failures
> Hudson monitors key health metrics of slaves
 Low disk space, insufficient swap
 Clock out of synch
 Extensible
> Slaves go offline automatically
> Catch problems before they break builds
23

Load Statistics Monitoring
24

When it’s time to add more slaves
25

Hudson made this extensible
> Hudson detects excessive workload
> Hudson notifies plugins
> Plugins can provision more slaves
 … assuming that you have that infrastructure
26

27

Amazon EC2: The Good
> Pay as you go (10¢/h or so)
 Loads on Hudson tend to be spiky
> Programmable API
> Instances launch at machine-speed
> EC2 instances are forgetful
28

Amazon EC2: The Bad
> Your data is still inside your firewall
 Takes time to check out code
 … or to archive build artifacts
 Some data just can’t be moved
> EC2 instances are forgetful
> Can your tests run in parallel?
29

Hudson EC2 plugin
> Built on top of typica*
> What does it do?
 Automatically provisions slaves on EC2 on demand
 Picks the right AMI depending on demand
 Starts slave agent
 Shuts down unused instances
* http://code.google.com/p/typica/ 30

Hudson “Appliance” on EC2
> Run the master in the cloud too, if you like
 Hudson on stock OpenSolaris AMI
 Data stored persistently in Elastic Block Storage
 Dynamically expandable thanks to ZFS
 Online, too
> Packaged as a wizard
31

Hudson EC2 plugin usage
> Tell Hudson what AMIs you want to start
32

33

Conclusion
> CI is here to stay
 We’ll continue to push more workload to servers
> Hudson makes this easy for you
> Reap the benefit of a cluster in multiple ways
34

Resources
> http://hudson.dev.java.net/
> Support Subscription
 hudson@sun.com
35

Arun Gupta
http://blogs.sun.com/arungupta
http://hudson.dev.java.net/
36