Package com.atlassian.jira.index.ha

Class DefaultIndexRecoveryManager

java.lang.Object

com.atlassian.jira.index.ha.DefaultIndexRecoveryManager

All Implemented Interfaces:

IndexRecoveryManager, Sized

public class DefaultIndexRecoveryManager
extends Object
implements IndexRecoveryManager

Field Summary

Fields

Modifier and Type	Field	Description
static final String	INDEX_FIXER

Fields inherited from interface com.atlassian.jira.index.ha.IndexRecoveryManager

INDEXSNAPSHOTS, SNAPSHOT_ARCHIVE

Constructor Summary

Constructors

Constructor

Description

DefaultIndexRecoveryManager(IndexLifecycleManager indexLifecycleManager, IndexPathManager indexPathManager, SharedEntityIndexManager sharedEntityIndexManager, IndexingCounterManager indexingCounterManager, IssueIndexManager indexManager, CommentManager commentManager, WorklogManager worklogManager, EntityVersioningManager entityVersioningManager, VersioningCleanupService versionsCleaner, com.atlassian.event.api.EventPublisher eventPublisher, ClusterNodes clusterNodes, IssueIndexer issueIndexer, InternalIndexingService internalIndexingService, I18nHelper i18nHelper)

Method Summary

Modifier and Type	Method	Description
IndexRecoveryResult	deindexEntitiesDeletedInTheLast(Duration duration, TaskProgressSink taskProgressSink)	Applies de-index based on the versioning table (delete column.* Why this method may be needed? It is possible in Jira (both server and DC) that a late "update" will override a "delete".
com.atlassian.core.util.DateUtils.DateRange	getDurationToCatchUpUsingVersions()	Gets the updated date range for issues that need to be re-indexed because the latest updates are missing from the index.
boolean	isEmpty()
static Duration	padDurationBetween(Instant priorTime, Supplier<Instant> currentTime)	Creates a Duration of length priorTime -> currentTime PLUS an extra 24 hours.
static String	readableDuration(Duration duration)
IndexCommandResult	recoverIndexFromBackup(File recoveryFile, TaskProgressSink taskProgressSink)	Recovers an index from an index backup.
IndexRecoveryResult	reindexWithVersionCheckEntitiesUpdatedInTheLast(Duration duration, TaskProgressSink taskProgressSink)	This method uses versioning to short-circuit building documents when the index version is already up-to-date with what's in the db.
IndexCommandResult	safeRecoverIndexFromBackup(File recoveryFile, TaskProgressSink taskProgressSink)	Recovers an index from an index backup.
int	size()

Methods inherited from class java.lang.Object

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

Field Details

INDEX_FIXER

public static final String INDEX_FIXER

See Also:

• Constant Field Values

Constructor Details

DefaultIndexRecoveryManager

public DefaultIndexRecoveryManager(IndexLifecycleManager indexLifecycleManager,
 IndexPathManager indexPathManager,
 SharedEntityIndexManager sharedEntityIndexManager,
 IndexingCounterManager indexingCounterManager,
 IssueIndexManager indexManager,
 CommentManager commentManager,
 WorklogManager worklogManager,
 EntityVersioningManager entityVersioningManager,
 VersioningCleanupService versionsCleaner,
 com.atlassian.event.api.EventPublisher eventPublisher,
 ClusterNodes clusterNodes,
 IssueIndexer issueIndexer,
 InternalIndexingService internalIndexingService,
 I18nHelper i18nHelper)

Method Details

safeRecoverIndexFromBackup

public IndexCommandResult safeRecoverIndexFromBackup(File recoveryFile,
 TaskProgressSink taskProgressSink)
                                              throws IndexException

Description copied from interface: IndexRecoveryManager

Recovers an index from an index backup. This method throws and stops processing if it's not able to calculate the duration where catch up needs to be applied.

Specified by:

safeRecoverIndexFromBackup in interface IndexRecoveryManager

Parameters:

recoveryFile - The backup file

taskProgressSink - The sink to report progress back.

Throws:

IndexException - If we are unable to recover the index or catch up after recovery is successful

recoverIndexFromBackup

public IndexCommandResult recoverIndexFromBackup(File recoveryFile,
 TaskProgressSink taskProgressSink)
                                          throws IndexException

Description copied from interface: IndexRecoveryManager

Recovers an index from an index backup. Unlike the IndexRecoveryManager.safeRecoverIndexFromBackup(java.io.File, com.atlassian.jira.task.TaskProgressSink) this method won't throw exception when not being able to to calculate the duration where catch up needs to be applied. It will skip the catch up and continue with other steps.

Specified by:

recoverIndexFromBackup in interface IndexRecoveryManager

Parameters:

recoveryFile - The backup file

Throws:

IndexException - If we are unable to recover the index or catch up after recovery is successful

reindexWithVersionCheckEntitiesUpdatedInTheLast

public IndexRecoveryResult reindexWithVersionCheckEntitiesUpdatedInTheLast(Duration duration,
 TaskProgressSink taskProgressSink)
                                                                    throws IndexException

Description copied from interface: IndexRecoveryManager

This method uses versioning to short-circuit building documents when the index version is already up-to-date with what's in the db.

Ensures that issues which have been modified between now and duration ago are up-to-date in the index. They are reindexed/deindexed as needed to achieve this.

Specified by:

reindexWithVersionCheckEntitiesUpdatedInTheLast in interface IndexRecoveryManager

Parameters:

duration - specifies how far back (from present time) to look for updated issues

taskProgressSink - A progress sink.

Returns:

Report how much time the operation took and how many issues were affected

Throws:

IndexException

deindexEntitiesDeletedInTheLast

public IndexRecoveryResult deindexEntitiesDeletedInTheLast(Duration duration,
 TaskProgressSink taskProgressSink)
                                                    throws IndexException

Description copied from interface: IndexRecoveryManager

Applies de-index based on the versioning table (delete column.* Why this method may be needed? It is possible in Jira (both server and DC) that a late "update" will override a "delete". To handle this problem, Jira has to replay de-index operations after "some time". See: EntityVersioningManager) See: VersioningDao See: NodeReindexService See: DefaultServerIndexRepairService

Specified by:

deindexEntitiesDeletedInTheLast in interface IndexRecoveryManager

Parameters:

duration - specifies how far back (from present time) to look for deleted issues

taskProgressSink - A progress sink.

Returns:

result of de-indexing

Throws:

IndexException

getDurationToCatchUpUsingVersions

public com.atlassian.core.util.DateUtils.DateRange getDurationToCatchUpUsingVersions()

Description copied from interface: IndexRecoveryManager

Gets the updated date range for issues that need to be re-indexed because the latest updates are missing from the index. The range is

from

Latest date in the index

to

Latest date in the database

The range can actually be negative, indicating there are indexed updates which do not belong and so may need to be removed.
Note: the latest db date is based on the issue_version table. If no versions are present, we fall back to the issues table (although deletes will not be reflected here).

Specified by:

getDurationToCatchUpUsingVersions in interface IndexRecoveryManager

Returns:

Date range of issues that need to be re-indexed.

padDurationBetween

public static Duration padDurationBetween(Instant priorTime,
 Supplier<Instant> currentTime)

Creates a Duration of length priorTime -> currentTime PLUS an extra 24 hours. It's padded is just in case there's a TZ discrepancy with DB vs JVM time. See @DBR-159. As we skip reindexing when the index version is already up-to-date, the penalty for the extra input is minimal.

readableDuration

public static String readableDuration(Duration duration)

size

public int size()

Specified by:

size in interface Sized

isEmpty

public boolean isEmpty()

Specified by:

isEmpty in interface Sized