com.unboundid.directory.sdk.sync.api

Class SyncSource

java.lang.Object

com.unboundid.directory.sdk.sync.api.SyncSource

All Implemented Interfaces:

Configurable, ExampleUsageProvider, UnboundIDExtension

@Extensible
 @SynchronizationServerExtension(appliesToLocalContent=false,
                                appliesToSynchronizedContent=true)
public abstract class SyncSource
extends java.lang.Object
implements UnboundIDExtension, Configurable, ExampleUsageProvider

This class defines an API that must be implemented by extensions in order to synchronize data from a generic (non-LDAP and non-JDBC) endpoint. Since the UnboundID Synchronization Server is LDAP-centric, this API allows you to take generic content and convert it into LDAP entries which can then be processed by the Synchronization Server. The lifecycle of a sync operation is as follows:

1. Detect change at the synchronization source
2. Fetch full source entry
3. Perform any mappings and compute the equivalent destination entry
4. Fetch full destination entry
5. Diff the computed destination entry and actual (fetched) destination entry
6. Apply the minimal set of changes at the destination to bring it in sync

This implies that the fetchEntry(SyncOperation) method will be called once for every change that is returned by getNextBatchOfChanges(int, AtomicLong).

This is a generic interface and there is no protocol-specific connection management provided. It is expected that implementers will provide their own libraries for talking to the source endpoint and handle the connection lifecycle in the initializeSyncSource(SyncServerContext, SyncSourceConfig, ArgumentParser) and finalizeSyncSource() methods of this extension.

During realtime synchronization (i.e. when a Sync Pipe is running), there is a sliding window of changes being processed, and this API provides a distinction between some different points along that window:

• Old changes: These are changes that the Sync Server has processed and acknowledged back to the Sync Source. The Sync Source is under no obligation to re-detect these changes.
• Startpoint: This marks where the Sync Source will start detecting changes if it is restarted.
• Detected but unacknowledged: These changes have been returned by getNextBatchOfChanges() but not completely processed and acknowledged back to the Sync Source.
• Undetected changes: The next call to getNextBatchOfChanges() should return the first changes that have not been detected. This should be somewhere at or ahead of the startpoint.

Several of these methods throw EndpointException, which should be used in the case of any connection or endpoint error. For other types of errors, runtime exceptions may be used (IllegalStateException, NullPointerException, etc.). The Synchronization Server will automatically retry operations that fail, up to a configurable amount of attempts. The EndpointException class allows you to specify a retry policy as well.

Configuring Sync Sources

In order to configure a Sync Source based on this API and written in Java, use a command like:

      dsconfig create-sync-source \
           --source-name "{source-name}" \
           --type third-party \
           --set "extension-class:{class-name}" \
           --set "extension-argument:{name=value}"
 

where "{source-name}" is the name to use for the Sync Source instance, "{class-name}" is the fully-qualified name of the Java class written using this API, and "{name=value}" represents name-value pairs for any arguments to provide to the sync source. If multiple arguments should be provided to the sync source, then the "--set extension-argument:{name=value}" option should be provided multiple times.

Constructor Summary

Constructors

Constructor and Description
SyncSource() Creates a new instance of this Sync Source.

Method Summary

Modifier and Type	Method and Description
abstract void	acknowledgeCompletedOps(java.util.LinkedList<SyncOperation> completedOps) Provides a way for the Synchronization Server to acknowledge back to the script which sync operations it has processed.
void	defineConfigArguments(com.unboundid.util.args.ArgumentParser parser) Updates the provided argument parser to define any configuration arguments which may be used by this extension.
abstract com.unboundid.ldap.sdk.Entry	fetchEntry(SyncOperation operation) Return a full source entry (in LDAP form) from the source, corresponding to the ChangeRecord that is passed in through the SyncOperation.
void	finalizeSyncSource() This hook is called when a Sync Pipe shuts down, when the resync process shuts down, or when the set-startpoint subcommand (from the realtime-sync command line tool) is finished.
abstract java.lang.String	getCurrentEndpointURL() Return the URL or path identifying the source endpoint from which this extension is transmitting data.
java.util.Map<java.util.List<java.lang.String>,java.lang.String>	getExamplesArgumentSets() Retrieves a map containing examples of configurations that may be used for this extension.
abstract java.lang.String[]	getExtensionDescription() Retrieves a human-readable description for this extension.
abstract java.lang.String	getExtensionName() Retrieves a human-readable name for this extension.
abstract java.util.List<ChangeRecord>	getNextBatchOfChanges(int maxChanges, java.util.concurrent.atomic.AtomicLong numStillPending) Return the next batch of change records from the source.
abstract java.io.Serializable	getStartpoint() Gets the current value of the startpoint for change detection.
void	initializeSyncSource(SyncServerContext serverContext, SyncSourceConfig config, com.unboundid.util.args.ArgumentParser parser) This hook is called when a Sync Pipe first starts up, when the resync process first starts up, or when the set-startpoint subcommand is called from the realtime-sync command line tool.
void	listAllEntries(java.util.concurrent.BlockingQueue<ChangeRecord> outputQueue) Gets a list of all the entries in the source endpoint.
void	listAllEntries(java.util.Iterator<java.lang.String> inputLines, java.util.concurrent.BlockingQueue<ChangeRecord> outputQueue) Gets a list of all the entries in the source from a given file input.
abstract void	setStartpoint(SetStartpointOptions options) This method should effectively set the starting point for synchronization to the place specified by the options parameter.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

SyncSource

public SyncSource()

Creates a new instance of this Sync Source. All sync source implementations must include a default constructor, but any initialization should generally be done in the initializeSyncSource(com.unboundid.directory.sdk.sync.types.SyncServerContext, com.unboundid.directory.sdk.sync.config.SyncSourceConfig, com.unboundid.util.args.ArgumentParser) method.

Method Detail

getExtensionName

public abstract java.lang.String getExtensionName()

Retrieves a human-readable name for this extension.

Specified by:

getExtensionName in interface UnboundIDExtension

Returns:

A human-readable name for this extension.

getExtensionDescription

public abstract java.lang.String[] getExtensionDescription()

Retrieves a human-readable description for this extension. Each element of the array that is returned will be considered a separate paragraph in generated documentation.

Specified by:

getExtensionDescription in interface UnboundIDExtension

Returns:

A human-readable description for this extension, or null or an empty array if no description should be available.

getExamplesArgumentSets

public java.util.Map<java.util.List<java.lang.String>,java.lang.String> getExamplesArgumentSets()

Retrieves a map containing examples of configurations that may be used for this extension. The map key should be a list of sample arguments, and the corresponding value should be a description of the behavior that will be exhibited by the extension when used with that configuration.

Specified by:

getExamplesArgumentSets in interface ExampleUsageProvider

Returns:

A map containing examples of configurations that may be used for this extension. It may be null or empty if there should not be any example argument sets.

defineConfigArguments

public void defineConfigArguments(com.unboundid.util.args.ArgumentParser parser)
                           throws com.unboundid.util.args.ArgumentException

Updates the provided argument parser to define any configuration arguments which may be used by this extension. The argument parser may also be updated to define relationships between arguments (e.g., to specify required, exclusive, or dependent argument sets).

Specified by:

defineConfigArguments in interface Configurable

Parameters:

parser - The argument parser to be updated with the configuration arguments which may be used by this extension.

Throws:

com.unboundid.util.args.ArgumentException - If a problem is encountered while updating the provided argument parser.

initializeSyncSource

public void initializeSyncSource(SyncServerContext serverContext,
                                 SyncSourceConfig config,
                                 com.unboundid.util.args.ArgumentParser parser)

This hook is called when a Sync Pipe first starts up, when the resync process first starts up, or when the set-startpoint subcommand is called from the realtime-sync command line tool. Any initialization of this sync source should be performed here. This method should generally store the SyncServerContext in a class member so that it can be used elsewhere in the implementation.

The default implementation is empty.

Parameters:

serverContext - A handle to the server context for the server in which this extension is running.

config - The general configuration for this sync source.

parser - The argument parser which has been initialized from the configuration for this sync source.

finalizeSyncSource

public void finalizeSyncSource()

This hook is called when a Sync Pipe shuts down, when the resync process shuts down, or when the set-startpoint subcommand (from the realtime-sync command line tool) is finished. Any clean up of this sync source should be performed here.

The default implementation is empty.

getCurrentEndpointURL

public abstract java.lang.String getCurrentEndpointURL()

Return the URL or path identifying the source endpoint from which this extension is transmitting data. This is used for logging purposes only, so it could just be a server name or hostname and port, etc.

Returns:

the path to the source endpoint (should not return null)

setStartpoint

public abstract void setStartpoint(SetStartpointOptions options)
                            throws EndpointException

This method should effectively set the starting point for synchronization to the place specified by the options parameter. This should cause all changes previous to the specified start point to be disregarded and only changes after that point to be returned by getNextBatchOfChanges(int, AtomicLong).

There are several different startpoint types (see SetStartpointOptions), and this implementation is not required to support them all. If the specified startpoint type is unsupported, this method should throw an UnsupportedOperationException.

IMPORTANT: The RESUME_AT_SERIALIZABLE startpoint type must be supported by your implementation, because this is used when a Sync Pipe first starts up. The Serializable in this case is the same type that is returned by getStartpoint(); the Sync Server persists it and passes it back in on a restart.

This method can be called from two different contexts:

• When the 'set-startpoint' subcommand of the realtime-sync CLI is used (the Sync Pipe is required to be stopped in this context)
• Immediately after a Sync Pipe starts up and a connection is first established to the source server (e.g. before the first call to getNextBatchOfChanges(int, AtomicLong))

Parameters:

options - an object which indicates where exactly to start synchronizing (e.g. the end of the changelog, specific change number, a certain time ago, etc)

Throws:

EndpointException - if there is any error while setting the start point

getStartpoint

public abstract java.io.Serializable getStartpoint()

Gets the current value of the startpoint for change detection. This is the "bookmark" which indicates which changes have already been processed and which have not. In most cases, a change number is used to detect changes and is managed by the Synchronization Server, in which case this implementation needs only to return the latest acknowledged change number. In other cases, the return value may correspond to a different value, such as the SYS_CHANGE_VERSION in Microsoft SQL Server. In any case, this method should return the value that is updated by acknowledgeCompletedOps(LinkedList).

This method is called periodically and the return value is saved in the persistent state for the Sync Pipe that uses this extension as its Sync Source.

IMPORTANT: The internal value for the startpoint should only be updated after a sync operation is acknowledged back to this script (via acknowledgeCompletedOps(LinkedList)). Otherwise it will be possible for changes to be missed when the Synchronization Server is restarted or a connection error occurs.

Returns:

a value to store in the persistent state for the Sync Pipe. This is usually a change number, but if a changelog table is not used to detect changes, this value should represent some other token to pass into setStartpoint(SetStartpointOptions) when the sync pipe starts up.

getNextBatchOfChanges

public abstract java.util.List<ChangeRecord> getNextBatchOfChanges(int maxChanges,
                                                                   java.util.concurrent.atomic.AtomicLong numStillPending)
                                                            throws EndpointException

Return the next batch of change records from the source. Change records are usually just hints that a change happened; they do not include the full contents of the target entry. In an effort to never synchronize stale data, the Synchronization Server will go back and fetch the full target entry for each change record.

On the first invocation, this should return changes starting from the startpoint that was set by setStartpoint(SetStartpointOptions). This method is also responsible for updating the internal state such that subsequent invocations do not return duplicate changes.

The resulting list should be limited by maxChanges. The numStillPending reference should be set to the estimated number of changes that haven't yet been retrieved from the source endpoint when this method returns, or zero if all the current changes have been retrieved.

IMPORTANT: While this method needs to keep track of which changes have already been returned so that it does not return them again, it should NOT modify the official startpoint. The internal value for the startpoint should only be updated after a sync operation is acknowledged back to this script (via acknowledgeCompletedOps(LinkedList)). Otherwise it will be possible for changes to be missed when the Synchronization Server is restarted or a connection error occurs. The startpoint should not change as a result of this method.

This method does not need to be thread-safe. It will be invoked repeatedly by a single thread, based on the polling interval set in the Sync Pipe configuration.

Parameters:

maxChanges - the maximum number of changes to retrieve

numStillPending - this should be set to the number of unretrieved changes that are still pending after this batch has been retrieved. This will be passed in as zero, and may be left that way if the actual value cannot be determined.

Returns:

a list of ChangeRecord instances, each corresponding to a single change at the source endpoint. If there are no new changes to return, this method should return an empty list.

Throws:

EndpointException - if there is any error while retrieving the next batch of changes

fetchEntry

public abstract com.unboundid.ldap.sdk.Entry fetchEntry(SyncOperation operation)
                                                 throws EndpointException

Return a full source entry (in LDAP form) from the source, corresponding to the ChangeRecord that is passed in through the SyncOperation. This method should perform any queries necessary to gather the latest values for all the attributes to be synchronized.

This method must be thread safe, as it will be called repeatedly and concurrently by each of the Sync Pipe worker threads as they process entries.

If the original ChangeRecord has the full entry already set on it (which can be done using ChangeRecord.Builder#fullEntry(Entry), then this method will not get called, and the Sync Server will automatically use the full entry from the ChangeRecord. In this case, the implementation can always return null.

Parameters:

operation - the SyncOperation which identifies the source "entry" to fetch. The ChangeRecord can be obtained by calling operation.getChangeRecord(). These ChangeRecords are generated by getNextBatchOfChanges(int, AtomicLong) or by listAllEntries(BlockingQueue).

Returns:

a full LDAP Entry, or null if no such entry exists.

Throws:

EndpointException - if there is an error fetching the entry

acknowledgeCompletedOps

public abstract void acknowledgeCompletedOps(java.util.LinkedList<SyncOperation> completedOps)
                                      throws EndpointException

Provides a way for the Synchronization Server to acknowledge back to the script which sync operations it has processed. This method should update the official startpoint which was set by setStartpoint(SetStartpointOptions) and is returned by getStartpoint().

IMPORTANT: The internal value for the startpoint should only be updated after a sync operation is acknowledged back to this extension (via this method). Otherwise it will be possible for changes to be missed when the Synchronization Server is restarted or a connection error occurs.

Parameters:

completedOps - a list of SyncOperations that have finished processing. The records are listed in the order they were first detected.

Throws:

EndpointException - if there is an error acknowledging the changes back to the database

listAllEntries

public void listAllEntries(java.util.concurrent.BlockingQueue<ChangeRecord> outputQueue)
                    throws EndpointException

Gets a list of all the entries in the source endpoint. This is used by the 'resync' command line tool. The default implementation throws a UnsupportedOperationException; subclasses should override if the resync functionality is needed.

The outputQueue should contain ChangeRecord objects with the ChangeType set to null to indicate that these are resync operations. ChangeRecords should be added to the queue using the blocking put() method.

This method should not return until all the entries at the source have been added to the output queue. Separate threads will concurrently drain entries from the queue and process them. The queue typically should not contain full entries, but rather ChangeRecord objects which identify the full source entries. These objects are then individually passed in to fetchEntry(SyncOperation). Therefore, it is important to make sure that the ChangeRecord instances contain enough identifiable information (e.g. primary keys) for each entry so that the entry can be found again.

The lifecycle of resync is similar to that of real-time sync, with a few differences:

1. Stream out a list of identifiers for the entries in the source endpoint, using a ChangeRecord as the identifier
2. Fetch full source entry for a ChangeRecord
3. Perform any mappings and compute the equivalent destination entry
4. Fetch full destination entry
5. Diff the computed destination entry and actual destination entry
6. Apply the minimal set of changes at the destination to bring it in sync

Alternatively, the full entry can be set on the ChangeRecord within this method, which will cause the "fetch full entry" step to be skipped. In this case the Sync Server will just use the entry specified on the ChangeRecord.

If the total set of entries is very large, it is fine to split up the work into multiple network queries within this method. The queue will not grow out of control because it blocks when it becomes full. The queue capacity is fixed at 1000.

Parameters:

outputQueue - a queue of ChangeRecord objects which will be individually fetched via fetchEntry(SyncOperation). ChangeRecords should be added to the queue using the blocking put() method.

Throws:

EndpointException - if there is an error retrieving the list of entries to resync

listAllEntries

public void listAllEntries(java.util.Iterator<java.lang.String> inputLines,
                           java.util.concurrent.BlockingQueue<ChangeRecord> outputQueue)
                    throws EndpointException

Gets a list of all the entries in the source from a given file input. This is used by the 'resync' command line tool. The default implementation throws a UnsupportedOperationException; subclasses should override if the resync functionality is needed for specific records, which can be specified in the input file.

The format for the inputLines (e.g. the content of the file) is user-defined; it may be key/value pairs, primary keys, or full SQL statements, for example. The use of this method is triggered via the --sourceInputFile argument on the resync CLI. The outputQueue should contain ChangeRecord objects with the ChangeType set to null to indicate that these are resync operations.

This method should not return until all the entries specified by the input file have been added to the output queue. Separate threads will concurrently drain entries from the queue and process them. The queue typically should not contain full entries, but rather ChangeRecord objects which identify the full source entries. These objects are then individually passed in to fetchEntry(SyncOperation). Therefore, it is important to make sure that the ChangeRecord instances contain enough identifiable information (e.g. primary keys) for each entry so that the entry can be found again.

The lifecycle of resync is similar to that of real-time sync, with a few differences:

1. Stream out a list of identifiers for entries in the source endpoint, using the given input file as the basis for which entries to resync
2. Fetch full source entry for an identifier
3. Perform any mappings and compute the equivalent destination entry
4. Fetch full destination entry
5. Diff the computed destination entry and actual destination entry
6. Apply the minimal set of changes at the destination to bring it in sync

Alternatively, the full entry can be set on the ChangeRecord within this method, which will cause the "fetch full entry" step to be skipped. In this case the Sync Server will just use the entry specified on the ChangeRecord.

If the total set of entries is very large, it is fine to split up the work into multiple network queries within this method. The queue will not grow out of control because it blocks when it becomes full. The queue capacity is fixed at 1000.

Parameters:

inputLines - an Iterator containing the lines from the specified input file to resync (this is specified on the CLI for the resync command). These lines can be any format, for example a set of primary keys, a set of WHERE clauses, a set of full SQL queries, etc.

outputQueue - a queue of ChangeRecord objects which will be individually fetched via fetchEntry(SyncOperation). ChangeRecords should be added to the queue using the blocking put() method.

Throws:

EndpointException - if there is an error retrieving the list of entries to resync