org.apache.giraph.block_app.framework

Interface BlockFactory<S>

Type Parameters:

S - Execution stage type

All Known Implementing Classes:

AbstractBlockFactory, MigrationFullBlockFactory, SimpleMigrationMasterBlockFactory

public interface BlockFactory<S>

Class describing a particular application. Everything except input and output should be fully encapsulated within this class. For any application, it should be enough to only specify particular BlockFactory. Given configuration, it creates a block that represents a full Giraph job. Recommended is to extend AbstractBlockFactory directly for most cases.

Method Summary

Modifier and Type	Method and Description
Block	createBlock(GiraphConfiguration conf) Create a block (representing a full Giraph job), based on the given configuration.
S	createExecutionStage(GiraphConfiguration conf) Create an empty instance of execution stage object.
List<String>	getGcJavaOpts(org.apache.hadoop.conf.Configuration conf) Get special GC Java options.
void	initConfig(GiraphConfiguration conf) Based on provided configuration, updates it, such that all necessary properties are initialized.
void	registerOutputs(GiraphConfiguration conf) Register outputs to use during the application (vs output at the end of the application), based on provided configuration.

Method Detail

initConfig

void initConfig(GiraphConfiguration conf)

Based on provided configuration, updates it, such that all necessary properties are initialized.

createBlock

Block createBlock(GiraphConfiguration conf)

Create a block (representing a full Giraph job), based on the given configuration. Configuration should be treated as immutable at this point. If there are issues in configuration, it is very cheap to throw from this method - as Giraph job will not even start. This function will be called two times - once before starting of the Giraph job, to fail early if anything is incorrectly configured. Second time will be on Master, which will return Block instance on which createIterator will be called once, which should return current application run. initConfig will be called only once, before starting Giraph job itself. Master will contain configuration already modified by initConfig.

createExecutionStage

S createExecutionStage(GiraphConfiguration conf)

Create an empty instance of execution stage object. Can be used by application to be aware of what was executed before. Most common example is counting iterations, or for having a boolean whether some important event happened. Execution stage should be immutable object, with creating a new object when different value is needed.

getGcJavaOpts

List<String> getGcJavaOpts(org.apache.hadoop.conf.Configuration conf)

Get special GC Java options. If returns null, default options are used.

registerOutputs

void registerOutputs(GiraphConfiguration conf)

Register outputs to use during the application (vs output at the end of the application), based on provided configuration.