Revision as of 11:16, 28 June 2010

Preliminary definitions

gCube resources and profiles

A gCube Resource is anything whose related information can be gathered, stored, monitored, and disseminated in order to provide the valuable amount of knowledge needed during the creation and management of a VRE as well as to operate an entire gCube infrastructure. In order to be appropriately managed and discovered, a gCube resource has to be described by creating a profile document compliant with its XML schema. The following resources have been identified:

• gHN (XSD)
• Service (XSD)
• Running Instance (XSD)
• External Running Instance (XSD)
• VRE (XSD)
• Collection (XSD)
• Metadata Collection (XSD)
• gLite resource (XSD)

Such resources can be combined or created at VRE creation time in order to set up a new Virtual Research Environment.

The gHN resource

A gCube Hosting Node (shortly, a gHN) is the resource modelling a node of an infrastructure able to host gCube Running Instances.

The gHN resource is built by collecting and grouping both the configuration and runtime information on the node. The profile of the gHN is composed by two classes of information:

• static information, i.e. information that does not change once the gHN is started

• dynamic information, i.e. information that changes during the gHN lifetime

At the top level, there are four main sections in the gHN profile:

• Infrastructure, reporting the name of the infrastructure to which the gHN is joined to

• GHNDescription, a GLUE compliant section describing the hardware characteristics of the node

• Site, reporting information about the site to which the gHN belongs to

• DeployedPackages, reporting the list of gCube software packages actually deployed on the gHN

[TBC]

The Service resource

Usually, a Service is defined as a software system that delivers functionalities. In gCube, a Service is a not-empty set of related Packages (connected through dependencies) forming an unique logical entity. A Package is the smallest installable unit of software that can be deployed on a gHN (e.g. a JAR o a GAR archive). Packages are the way in which the software needed to set up a VRE has to prepared in order to be used and stored in the system. Once the service components have been developed, they must be described by compiling the service profile.

Package types

Concretely, a Package is a “piece of software” that can be deployed on a gHN. A Package can be:

• a MainPackage, representing a package that once deployed produces a WSRF Service (a service able to manage stateful resources following the WS-Resource patterns);
• a Software, a software library (typically composed by one or more JAR archives) offering functionality for interacting with Service or a third party software distributed with the service that can be dynamically deployed;

About how to define a package, see the Packages section.

Composition

The set of Packages forming a Service is composed by:

1. one and only one MainPackage representing the gCore compliant Web-Service
2. an arbitrary number of other Packages of different types logically related (even if not used) to the MainPackage

The "main" part of a Service is, of course, the MainPackage from which the related Running Instance resource is generated. The other packages can be:

• helper modules developed and used by the MainPackage i.e stubs
• helper modules developed and distributed with the MainPackage in order to be used by other Services to interact with a Running Instance of the Service. e.g. test-suite

We enforce the use of this convection:

Main Package Name: ServiceName-service

Stubs: ServiceName-stubs

Test suite: ServiceName-test-suite

Example if the ServiceName is Deployer, we will use:

Main Package Name: Deployer-service

Stubs: Deployer-stubs

Test suite: Deployer-test-suite

How to behave with library

A library can be profiled in several ways, mainly depending of its source. The following table depicts them:

Example of profile for D4Science library not service specific

This is a basic template for a D4Science library profile:

 <Resource>
 	<ID></ID>
 	<Type>Service</Type>
 	<Profile>
 		<Description>....</Description> 
 		<Class>... a class name...</Class>
 		<Name>... a library name...</Name>
 		<Version>\d{1,2}+.\d{1,2}+.\d{1,2}</Version>
 		…
 		<Packages>
 			<Software>
 				<Description></Description>
 				<Name>... a library name...</Name>
 				<Version>\d{1,2}+.\d{1,2}+.\d{1,2}</Version>
 				…
 			</Software>
 		</Packages>
 	</Profile>
 </Resource>

where:

... a class name... has to be replaced with the name of Service Class the library belong to

... a library name... has to be replaced with the name of Library name

\d{1,2}+.\d{1,2}+.\d{1,2} has to be replaced with the actual version of the Library

The following is an instance example of the template:

 <Resource>
 	<ID></ID>
 	<Type>Service</Type>
 	<Profile>
 		<Description>....</Description>
 		<Class>Search</Class>
 		<Name>ResultSetLibrary</Name>
 		<Version>1.0.0</Version>
 		<Packages>
 			<Software>
 				<Description>RS Library</Description>
 				<Name>ResultSetLibrary</Name>
 				<Version>1.0.0</Version>
 				...
 			</Software>
 		</Packages>
 	</Profile>
 </Resource>

Example of profile for Service Independent (Deployable) library

This is a basic template for a Service Independent (Deployable) library profile:

 <Resource>
 	<ID></ID>
 	<Type>Service</Type>
 	<Profile>
 		<Description></Description> 
 		<Class>ExternalSoftware</Class>
 		<Name>... a library name...</Name>
 		<Version>\d{1,2}+.\d{1,2}+.\d{1,2}</Version>
 		…
 	<Packages>
 		<Software>
 			<Description></Description>
 			<Name>... a library name...</Name>
 			<Version>\d{1,2}+.\d{1,2}+.\d{1,2}</Version>
 			…
 		</Software>
 	</Packages>
 </Resource>

where

... a library name... has to be replaced with the name of Library name

\d{1,2}+.\d{1,2}+.\d{1,2} has to be replaced with the actual version of the Library

The following is an instance example of the template:

 <Resource>
         <ID></ID>
         <Type>Service</Type>
         <Profile>
                 <Description>Commons HTTP Client library</Description>
                 <Class>ExternalSoftware</Class>
                 <Name>commons-httpclient</Name>
                 <Version>3.0.1</Version>
                 <Packages>
                         <Software deployable="true">
                                 <Description>Commons HTTP Client library</Description>
                                 <Name>commons-httpclient</Name>
                                 <Version>3.0.1</Version>
                                 ...
                         </Software>
                 </Packages>
                 <SpecificData></SpecificData>
         </Profile>
 </Resource>

How to specify dependencies

A package, once deployed, can rely on other software to perform its activities. Let's see how to specify these dependencies by distinguishing among:

• software developed within this service (this is the case of a stub library, for instance)

  → the software has to be declared as Software and the dependency is specified as package Dependency

• software developed and distributed within other gCube services

  → the dependency is specified as package Dependency against a package defined in another Service profile

• third party deployable software used only by this service

  → the software has to be declared as Software and the dependency is specified as package Dependency

• third party deployable software used by other services

  → the software has to be uploaded in the ETICS repository, the package has to be declared as Software and the dependency is specified as package Dependency

• third party software that is not dynamically deployable

  → the software has to be manually installed, the label has to be manually declared in the gHN profile (by the gHN owner) and the dependency is specified as gHN Requirement

How to compile a Service profile

The following example shows the common elements that compose the Service profile. The Package-specific sections are explained apart later.

<Resource xmlns:wsdl="http://schemas.xmlsoap.org/wsdl/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
	<UniqueID/>
	<Type>Service</ResourceType>
	<Scopes>
	</Scopes>
	<Profile>
		<Description> ...a description...</Description>
		<Class> ... a class name... </Class>
		<Name>...a service name...</Name>
		<Version>\d{1,2}+.\d{1,2}+.\d{1,2}</Version>		
		<Dependencies>
			<Dependency>
				<Class>...</Class>
				<Name>...</Name>
				<Version>\d{1,2}+.\d{1,2}+.\d{1,2}</Version>
			</Dependency>
			<Dependency>
				<Class>...</Class>
				<Name>...</Name>
				<Version>\d{1,2}+.\d{1,2}+.\d{1,2}</Version>
			</Dependency>
			<Dependency>
				<Class>...</Class>
				<Name>...</Name>
				<Version>\d{1,2}+.\d{1,2}+.\d{1,2}</Version>
			</Dependency>
		</Dependencies>
		<SpecificData/>
		<Packages>
			[...]
		</Packages>
	</Profile>
</Resource>

Common elements

Class

The class of a service is its functional area. It's a free text field that should group packages belonging the same logical functionalities.

Name

The Name is an arbitrary string that uniquely identifies the service within a service class.

Version

The service version must be in the format: \d{1,2}+.\d{1,2}+.\d{1,2}

This version marks a whole gCube distribution and, for the current distribution under development in D4Science, the version number is fixed to 1.00.00. Future and different distributions will eventually evolve with a different service version.

Dependencies

A gCube service is part of a complex system and it makes use of other services. At VRE definition time, firstly, it is compiled the list of services that satisfy the VRE definition criteria and, then, such a list is completed with the other services that allow the first ones to work properly. Therefore it is needed that a service declares which other services it will need to be available in order to do its work once it is deployed.

Packages section

This element groups an unbounded sequence of Package-derived elements. Each element describes a package that comes with the service. The structure of a Package element includes an initial common part and a package-specific part rooted by an element named as the service type (MainPackage and Software). The only constraint is that one and only one MainPackage element is accepted in the list.

The following example shows the elements of a Package definition common to both types of Package.

 <Packages>
 	<Main deployable="true">
 		<Name>Deployer-service</Name>
 		<Version>1.0.0</Version>
 		<Mandatory level="GHN"/>
 		<Shareable level="VO"/>
 		<GHNRequirements>
 			<Requirement category="Site"  requirement="string" value="java1.5" operator="ge"/>
 		</GHNRequirements>
 		<Dependencies>
 			<Dependency>
 				<Service>
 					<Class>VREManagement</Class>
 					<Name>Deployer</Name>
 					<Version>1.0.0</Version>
 				</Service>
 				<Package>Deployer-stubs</Package>
 				<Version>1.0.0</Version>
 				<Scope level="GHN"/>
 				<Optional>false</Optional>
 			</Dependency>
 			<Dependency>
 				<Service>
 					<Class>VREManagement</Class>
 					<Name>SoftwareRepository</Name>
 					<Version>1.0.0</Version>
 				</Service>
 				<Package>SoftwareRepository-stubs</Package>
 				<Version>1.0.4</Version>
 				<Scope level="GHN"/>
 				<Optional>false</Optional>
 			</Dependency>
 		</Dependencies>
 		<GARArchive>org.gcube.common.vremanagement.deployer.gar</GARArchive>
 		<PortType>
 			<Name>gcube/common/vremanagement/Deployer</Name>
 			<Security/>			
 			<WSDL>
 			</WSDL>
 		</PortType>					
 	</Main>
 </Packages>

Name

An arbritrary string that uniquely identifies the package within this service.

Version

The version of the package. It has to be in the format: \d{1,2}+.\d{1,2}+.\d{1,2} (e.g. 1.0.1 or 1.11.0)

Mandatory

If the Mandatory element is declared, it states when the package has to be deployed by default. The value of the attribute level defines at which level the package is mandatory. Accepted values are: NONE, GHN, VO and VRE. For instance, stating that a package is mandatory at VO level means that each time a new VO is created the package MUST be deployed.

If the element is not declared, the package is assumed to be not mandatory at any level by default.

Shareable

If the Shareable element is declared, it states if the package can be shared across multiple Scopes. The value of the attribute level defines at which level the package is shareable. Accepted values are: NONE, VO and VRE.

If the value is NONE, the package cannot be shared across any scope, i.e. there must be one and only one deployed instance of that package in a scope.

If the element is not declared, the package is assumed to be shareable at any scope level by default.

InstallScripts

This set of scripts are executed before the deployment of the package. They can be used to prepare the environment for the package execution (create a file system structure, install third parties software). The current folder where they are executed is the one where it is places. So, one can navigate the package tree with relative paths starting from the install script folder.

UninstallScripts

This set of scripts are executed after the undeployment of the package. They can be used to clean up the environment. The current folder where they are executed is the one where the package is downloaded and uncompressed.

RebootScripts

This set of scripts are executed before each container startup.

Dependencies

The dependencies of the package are a list of other packages that allow the package to properly work. These packages *must* defined somewhere in the same profile (and in this case they have the same ServiceClass and ServiceName of the package) or in other profiles. It is important to point out that also the dependencies against the packages declared in the current profile has to be specified (like the one against its own stubs library).

A dependency is a declaration of the following elements:

• Service
• PackageName
• Version

Need to be able to declare minimum, maximum allowed versions of a dependency (both min and max may be optional), and allow "holes" for known incompatible versions.

Mathematical syntax chosen to avoid the use of - as it would conflict with what is used in many version number, and because < and >= doesn't look nice in XML. (,1.0.0] is used because infinity is not really helpful here.

Default strategy: Of the overlapping ranges, the highest soft requirement is the version to be used. If there are no soft requirements inside the prescribed ranges, the most recent version is used. If that does not fit the described ranges, then the most recent version number in the prescribed ranges is used. If the ranges exclude all versions, an error occurs.

Addition of ranges leads to additional necessary specifications on the dependency element

• Scope level

"GHN" means that the package must be co-deployed on the same GHN

"VRE" means that the package must be deployed on a GHN and visible in the same VRE (typically used for dependencies against MainPackages)

"VO" means that the package must be deployed on a GHN and visible in the same VO (typically used for dependencies against MainPackages)

• Optional

a false value means that the fulfilment of dependency is mandatory for a successful deployment of the package that declares the dependency

The example above defines three dependencies that requires that the stub classes of the Deployer service and the stub classes of the SoftwareRepository service, have to be co-deployed with the Deployer-service package.

GHNRequirements

A GHNRequirement is a requirement against the node where the package is going to be deployed. These requirements are matched with the gHN profiles in order to find a suitable node that can host the package. A gHN requirement is expressed by:

• category: it identifies the element/attribute of the Profile/GHNDescription section of the gHN profile against which the requirement is expressed.

It can assume one of the following values:

• operator: declares the relation between the actual value of the looked up attribute category and the value expressed in the Requirement.

It can assume one of the following values:

• requirement: one of the attribute and/or child elements defined for the selected category in the gHN profile

• value: the value of the selected attribute defined for the category in the GHN profile

Therefore, a gHN requirement is an expression evaluated in this form: <category/requirement operator value>

• A simple example:

The example above means: "This package can be deployed on a gHN running on a Linux Operating System (Line 2) (it requires that the value of OS in the GHN profile contains the string Linux) and having at least 2 processors (Line 3)". Note: in both cases the requirement attribute is not relevant.

• Example. Accessing the run time environment variables:

The former block (Line 2) checks that the environment contains the key ANT_HOME and that its value is equal to /usr/share/ant while the latter (Line 3) simply checks that the environment contains the key GLOBUS_OPTIONS regardless its value.

• '''Putting them together a clarifying example''':

Here a concrete example where the user statically claims that its service to run needs:

• A jdk having version 1.5.* (note the exist operator)
• An installed gCore having version 1.0.1
• That apache tomcat is installed on the machine (it is enough to check the CATALINA_HOME environment variable is defined)
• The ghn must run in the *.research-infrastructures.eu domain
• The operating system must be Linux*
• The platform must be i386

• Example. Custom queries:

To give freedom to customize the access to specific paths in the resources a CUSTOM_REQUIREMENT category has been introduce. The example below is a sample:

The same result can be obtained by the query:

GHNRequiremnts can be also used when a package depends on a software that cannot be dynamically deployed. In this case, such a software has to be

• manually installed,
• declared in the gHN profile as a conventional RunTimeEnv/Variable value and,
• reported as GHNRequirement in the package definition.

The list of labels is open and new labels can be defined at any time. So far, the list includes:

• java1.5 - a JVM v1.5 is available
• gLiteSEAccess - the node is configured to access to a Storage Element on a gLite infrastructure
• MySQLdb - a MySQL database is available
• Oracle10g - an Oracle10g database is available

Package-specific sections

Depending on the type of the package, there are some specific elements to report on the package profile.

MainPackage

A MainPackage is a package than once deployed it generates an instance of a gCube Service.

For convention the main package name have to be in the form: ServiceName-service

Example:

 <Packages>
 	<Main deployable="true">
 		<Name>Deployer-service</Name>
 		<Version>1.0.0</Version>
 		<Mandatory level="GHN"/>
 		<Shareable level="VO"/>
 		<GHNRequirements>
 			<Requirement category="Site"  requirement="string" value="java1.5" operator="ge"/>
 		</GHNRequirements>
 		<Dependencies>
 			[...]
 		</Dependencies>
 		<GARArchive>org.gcube.common.vremanagement.deployer.gar</GARArchive>
 		<PortType>
 			<Name>gcube/common/vremanagement/Deployer</Name>
 			<Security/>			
 			<WSDL>
 			</WSDL>
 		</PortType>					
 	</Main>
 </Packages>

GARArchive

The full path of the GAR file. The path must start from the package folder.

Example: If the structure of the directory is like this

|--Deployer-service
   \--archive
      \--org.gcube.common.vremanagement.deployer.gar

The declaration in the profile must be:

 <GARArchive>archive/org.gcube.searchservice.resultsetservice.gar</GARArchive>

PortType

A PortType is a WSDL interface generated by the deployment of the MainPackage's GAR. Each of them is declared in the Deployment Descriptor file (WSDD) with a <service> element.

Name

The value of the name attribute in the <service> element in the service's WSDD file.

WSDL

A section to leave empty. It is filled at build time by the gCore utilities with the actual WSDL of the portType.

Software

A Software package is a deployable library or a third party package uploaded with the service.

Example:

 <Packages>
 	<Software>
 		<Name>Deployer-stubs</Name>
 		<Version>1.0.0</Version>
 		[...]			
 		<Files>
 			<File>org.gcube.common.vremanagement.deployer.stubs.jar</File>
 		</Files>
 	</Software>
 </Packages>

Files

The relative path (starting from software archive root directory) and the name of the library's files shipped in the software archive.