Configuration management - The Software Enterprise

Download Report

Transcript Configuration management - The Software Enterprise

Configuration Management (CM)*
Source code control
Versioning
Build Management
*Some of this material adapted from Ian Sommerville’s course notes, Ch. 29, and Wigerd, L. and Seiwald, 1998
Change Management
Change happens!
 Don’t get frustrated as to “why”, but understand that it
will happen for a variety of reasons:
• Sales uncovers a “can’t miss opportunity” if they just had one
more feature
• Customers and business analysts provide incomplete and
ambiguous requirements
• Technology foundation changes
• …and a zillion other reasons
Change Management processes identify what
needs to be changed and who it impacts
 We discuss at length in 416, for you what is important
is that many source code files may change in
response to one change event.
Configuration management (CM)
The development & application of procedures &
standards to manage an evolving software product
New versions are created as software evolves
 For different machines/OS
 Offering different functionality
 Tailored for particular user requirements
CM is concerned with managing evolving software
 Software change is a team activity
 CM aims to control the costs and effort involved in making
changes to a system
 May be seen as part of more general quality and change
management processes
CM and CM
abc.java
CHANGE
EVENT
CHANGE
EVENT
def.java
xyz.java
What happens when multiple changes affect the
same artifacts?
CM Concepts:Configurations & Releases
Configuration: An instance of a system composed
of specific versions of its artifacts
 Includes expectations of the target environment(s)
Release 2
abc.java v2.1
def.java v1.0
abc.java v2.0
abc.java v1.0
xyz.java v3.0
xyz.java v2.0
Release 1
xyz.java v1.0
Release: An instance of a system distributed to
users outside of the development team
 Releases may be targeted for (in)external communities
CM Concepts:Codelines & Branches
Codelines define a trajectory for a collection of
(source code) artifacts
 You have a history
 You have a notion of where it is going
 You have a set of policies governing participation
Branch - A branch is a named variant of a
codeline. That is
 a collection of software artifacts
 Assigned a logical identifier
 Whose purpose is to be either folded back into the
main codeline, or maintained as a release.
CM Concepts:Workspace & Repository
A source code control (SCC) repository is a
shared file system of software artifacts (server-side)
 Typically supported with client/server tools
 Typically understands versions and variant semantics of
the artifacts
 Often provides some mechanism for assigning jobs to
change control on software artifacts.
Workspaces are local (client) repositories
 Where developers build, test, & debug.
 Developers must periodically synch with SCC repository
CM Tools
A configuration management tool is one that
manages the acceptable set of configurations
targeted for different releases
A source code control tool is a repository that
manages the (source code) artifacts associated
with a project
An SCC repository is part of a larger CM system. CM
also includes tools and processes for building
deployable (executable) artifacts from the SCC
repository.
Branching/Merging (per artifact)
Exclusive locking – one codeline
Changes visible immediately
after check-in
Concurrent development using
branching – multiple codelines
Synchronize concurrent
development using merges – merge
between codelines
Changes private and not part of
mainline until merge – better
supports large, long changes
From http://www.cmcrossroads.com/bradapp/acme/branching/
CM Best Practices
Branching: Mainline model (yes) Promotion Model (NO!)
Branches: variants of the code artifacts.





Branch only when necessary.
Don’t copy when you mean to branch.
Branch on incompatible policy.
Branch late.
Branch, instead of freeze.
CM Best Practices*
Codelines:
 Give each codeline a policy.
 Give each codeline an owner.
 Have a mainline.
Change propagation: getting changes from one
codeline to another.
 Make original changes in the branch that has evolved
the least since branching.
 Propagate early and often.
 Get the right person to do the merge.
*Wingerd & Seiwald, 1998. See References
CM Best Practices*
Workspaces:





Don’t work outside of managed workspaces
Stay in sync with the codeline.
Don’t share workspaces.
Don’t use jello views.
Check in often.
SCC Repository
Local Repository
abc.java
abc.java
def.java
def.java
METADATA
*Wingerd & Seiwald, 1998. See References
Sandboxes and Vendor Branches
A couple of problems solved with these concepts:
1. You have an idea not fully fleshed out that you want to
prototype as part of a requirements understanding
activity. You want to put the code under SCC, but it
shouldn’t be part of the main product (yet, until the
requirements are validated)
Solution: Use a “Sandbox”, which is a separate
codeline with distinct quality (checkin) policies
2. You need to track dependencies to open source (or
COTS if you have the code) artifacts and manage
changes to this underlying platform.
Solution: use a vendor branch, which allows one to recheckin the new vendor version and resolve merge
conflicts
SCM Classes of Information*
Examples:






Introduction: What are the CI’s (Configurable Items)
Mgmt: Can be tricky in secure, multi-vendor, & IP-sensitive envs
Activities: The main “stuff” you do to make CM happen (next slide)
Schedules: Relationships to master project plan for change procs
Resources: Tools, repositories, guides, training, etc.
Plan Maintenance: Managing change to the CM process
*IEEE Standard 828-2005
SCM Activities
1. Configuration Identification: Identify, Name, Describe
what needs change control
2. Configuration Control: Identify need for change,
analyze, approve/disapprove, verify/release
3. Configuration Status Accounting: What metrics are to
be tracked?
• Number and frequency of changes on an item
4. Configuration Evaluation & Reviews: Essentially a
validation task - does the configured software
(baseline) meet requirements?
5. Release Mgmt & Delivery: Description of how software
is built and released to external users, and plans for
managing change post-release. (…stay tuned)
Recommended Reading
• Sommerville, Ian. Software Engineering, 7th ed., Chapter
29. Addison-Wesley, 2004.
• Wingerd, L. and Seiwald, C. “High-level Best Practices in
Software Configuration Management”, ECOOP 98, SCM8, Springer-Verlag LNCS 1439, 1998.
• Fogel, K. and Moshe, B. Open Source Development with
CVS (3rd ed.) Available at http://cvsbook.red-bean.com.
• IEEE Standard 828-2005, IEEE Standard for Software
Configuration Management Plans
• Software Engineering Body of Knowledge (SWEBOK),
Chapter 7, Software Configuration Management, 2004
System building
The process of compiling and assembling software
components into an executable system

Product lines built from combinations of components
abc.java
def.java
xyz.properties
xyz.xml
foo.exe
Invariably supported by automated tools that are
driven by ‘build scripts’

7/18/2015
i.e. Ant, Maven, Make, CMake
17
Example Build Issues
Do build scripts include all required components?
 When there are many hundreds of components
making up a system, it is easy to miss one. This is
normally detected by a build verification tool
Is the appropriate component version specified?
 A more significant problem. A system built with the
wrong version may work initially but fail after delivery
Are all configuration files available?
These and 1000+ other things can
make the process go bad
7/18/2015
18
System Build Process
Build Management System
(results, logs, notification)
Build
(compiler, etc.)
CM System
Build Script
(Ant-based)
Scripts
Scripts
Scripts
Source
Artifacts
Post-build
(deploy, validate)
Generated
Artifacts
Test results
Packaged for
deployment
•All activities provide status and results to Build Management System
•Build system executes build scripts against specific course code version
•Configuration identifies artifact versions for build
•Built tools generate and package deploy artifacts
•Post-build may include executing tests (JUnit) and deployment
System representation
Systems are normally represented for building by
using filenames to be processed by build tools.


Dependencies between files described to build tools
Mistakes can be made as users lose track of which
objects are stored in which files
It is useful to know the distinction between compiletime dependencies and run-time dependencies
abc.java
foo.exe
JDBC
7/18/2015
Mysql.jar
20
Common Build Goals
Ensure all required components are included
 Changes over time as system grow
Identify changes going into this build
 Tightly coupled with CM and change tracking
Ability to recreate the software version built
Share builds with the team including
 What was built – code version, which work items
 Logs, Test results, Build results
Build validation tasks depends on type of build
 Release– full validation,
• Complete testing – functional, perf/scale, code coverage, etc.
 Milestone – partial validation
• Most tests – no surprises before release
 Build – “smoke test”, a small, reliable subset of tests
• Goal is quick, yet reliable validation
...
Build Management Best Practices
1.
2.
3.
4.
5.
Define the process.
Know WHO owns the build & WHO is the customer.
Know WHAT the requirements for the build are
Know WHERE the build is targeted to go.
Number/ID your builds - use mnemonics that suggest the
source code version and platform
6. Change Management - know what changed from build to
build - code and build scripts
7. Validate the result – how is it accepted? What tests to run?
8. Track build process defects
9. Make external dependencies explicit - 3rd party packages,
configuration files, etc.
10.If possible, have developers use the build tools
7/18/2015
22
Jakarta Ant
Ant is a Java-based build tool
 Originally written by James Duncan Davidson on an
airplane from Europe to the US.
• As a Java developer, he tired of make’s poor cross-platform abilities
 Like ‘make’ for Java – only better
 Uses an XML file to drive actions
Benefits of Ant






Cross-platform, runs anywhere Java runs
Java domain smart (Junit, Eclipse)
Extremely powerful, modular, extensible
Fast, optimized (ex: javac run once for all subdirectories)
Extensible, strong community support
Just a few alternatives – maven, cernunnos
Ant Example – build.xml file
Target to execute if none specified
<project name="test" default="dist" basedir=".">
<property name="src.dir" value="."/>
<property name="build.dir" value="build"/>
<target name="init">
<mkdir dir="${build.dir}"/>
</target>
All activity relative
to basedir. Uses
build.xml file’s
directory if omitted
Properties are ant’s userdefined variables
Targets group a collection of
commands to execute
<target name="compile" depends="init">
<javac srcdir="${src.dir}" destdir="${build.dir}"/>
</target>
<target name="dist" depends="compile">
<jar basedir="${build.dir}" destfile="${ant.project.name}.jar" />
</target>
<target name="deploy" depends="dist">
<ftp server="10.1.1.1" userid="user" password="pass" />
</target>
</project>
Properties
Ant’s parameterize-ability
Using properties
 ${property.name} to use a property value in build file
Setting properties
 Name/value
<property name="project.home" value="/usr/project"/>
 Load properties from properties file
<property file="${project.home}/build.properties”/>
 Load environment variables
<property environment="env"/>
Useful Ant built-in properties:
 ant.project.name – set to <project>’s name attribute
 basedir – set to build file’s directory unless changed
in <project>
Ant Targets
Targets contain order tasks to execute when target
is invoked
 Targets can depend on other targets that must
execute before their execution
All targets have these attributes
 name (required)
• Common names – init, compile, clean
 depends (comma separated list of other targets to
execute first)
 description (for –projecthelp)
 if/unless (conditional execution)
<target name=”build.debug” if=”debug”>...</target>
<target name=”build.prod” unless=”debug”>...</target>unless
Useful Target Examples
init
 Set up properties, single definition for entire build file
 Can be set directly or in a property file
<target name="init">
<property name="name" value="test"/>
<property file="build.properties"/>
<!-- used later in build -->
<path id="compile.classpath">
<pathelement path="${weblogic.jar}"/>
</path>
</target>
Useful Target Examples
prepare
 Create directories needed for building/packaging
 Copy files to locations for packaging/deploying
<target name="prepare" depends="init">
<tstamp/> <!-- set a timestamp -->
<mkdir dir="${build.dir}/META-INF"/>
</target>
clean
 Removes all generated artifacts from build process
<target name="clean">
<delete dir="${build.dir}" />
<delete dir="gen-src" />
<delete file="${jarfile}" />
</target>
Useful Target Examples
compile
 Compiles code in source directories to destination directory
 Powerful classpath and classpath referencing
 Performs .java/.class out-of-date checking
<target name="compile" depends="prepare">
<javac
debug="true"
target="1.1"
source="1.2"
srcdir="${src.dir}"
destdir="${build.dir}"
excludes="share/ws/*, **/messaging/*">
<classpath refid="compile.classpath" />
</javac>
</target>
Useful Target Examples
dist
 Create any directory structures needed for
building/packaging
 Copy files to locations for packaging/deploying
<target name=“dist" depends="compile">
<jar
destfile="${basedir}/${ant.project.name}.jar"
basedir="${build.dir}"/>
</target>
Ant’s Types
Ant has several tags that function like data types
 PatternSet, FileSet, DirSet, Path-Like Structures
 Used by tasks to select groups of files
 All have an id attribute so they can be specified and
referenced later in the build file
PatternSet
 Filters files for selection in a command
<patternset id="xml.files" includes="**/*.xml"
excludes="**/*Test*"/>
<patternset id="xml.files">
<include name="**/*.xml"/>
<exclude name="**/*Test*"/>
</patternset>
Ant’s Types
FileSet/DirSet
 Selects a group of files/directories based on a pattern
 PatternSets can be nested in FileSets
 Used a lot in other tasks (jar, zip, javac) to select files
<fileset dir="${src.dir}" includes="**/*.java"/>
<fileset dir="${src.dir}">
<includes name="**/*.java"/>
</fileset>
Path-like structures
 Represents path items – classpath, path, etc.
 Ant handles platform-specific delimiters - : ; / \
<path id="test.classpath">
<path refid="classpath"/>
<pathelement path="${classes.dir}"/>
</path>
Ant Command-line Arguments
-help
-projecthelp
-quiet
-verbose / -debug
-logfile <file>
-D<property>=<value>
-buildfile/file/f <file>
-propertyfile <file>
… and others
display ant help message
displays description field for all targets
do not print information about build process
print more detailed information about build
write output to specified logfile
define a property and value
use designated file instead of ./build.xml
load all properties in designated file
Summary: Goals for CM/Build Process
Defined: You know what your process looks like
and where you are in it
Repeatable: If you take the same source, the same
configuration, and the same target platform, you
should get the exact same result
Verifiable: There is a defined and measurable
criteria for an acceptable build
Appreciated: In other words, planned. Your team
should not assume that because it compiles it
can be instantaneously built/deployed anywhere
Efficient: You do not want to drag the team to a halt.
7/18/2015
Build Management 2008
34