org.apache.ignite.internal.processors.cache.persistence.snapshot

Class IgniteSnapshotManager

java.lang.Object

org.apache.ignite.internal.processors.cache.GridCacheSharedManagerAdapter

org.apache.ignite.internal.processors.cache.persistence.snapshot.IgniteSnapshotManager

All Implemented Interfaces:

IgniteSnapshot, PartitionsExchangeAware, GridCacheSharedManager, MetastorageLifecycleListener, IgniteChangeGlobalStateSupport

public class IgniteSnapshotManager
extends GridCacheSharedManagerAdapter
implements IgniteSnapshot, PartitionsExchangeAware, MetastorageLifecycleListener, IgniteChangeGlobalStateSupport

Internal implementation of snapshot operations over persistence caches.

These major actions available:

• Create snapshot of the whole cluster cache groups by triggering PME to achieve consistency.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected static class	IgniteSnapshotManager.ClusterSnapshotFuture
protected static class	IgniteSnapshotManager.SnapshotHandlers Snapshot operation handlers.

Field Summary

Fields

Modifier and Type	Field and Description
static String	CP_SNAPSHOT_REASON Text Reason for checkpoint to start snapshot operation.
static String	DELTA_SUFFIX File with delta pages suffix.
static String	DFLT_SNAPSHOT_TMP_DIR Default snapshot directory for loading remote snapshots.
static long	DFLT_SNAPSHOT_TRANSFER_RATE_BYTES Snapshot transfer rate is unlimited by default.
static String	INDEX_DELTA_NAME File name template for index delta pages.
static String	PART_DELTA_TEMPLATE File name template consists of delta pages.
static String	RMT_SNAPSHOT_PREFIX Name prefix for each remote snapshot operation.
static int	SNAPSHOT_LIMITED_TRANSFER_BLOCK_SIZE_BYTES Maximum block size for limited snapshot transfer (64KB by default).
static String	SNAPSHOT_METAFILE_EXT Snapshot metafile extension.
static String	SNAPSHOT_METRICS Snapshot metrics prefix.
static String	SNAPSHOT_RUNNER_THREAD_PREFIX Prefix for snapshot threads.
static String	SNAPSHOT_TRANSFER_RATE_DMS_KEY Snapshot transfer rate distributed configuration key
static String	SNP_IN_PROGRESS_ERR_MSG Snapshot in progress error message.
static String	SNP_NODE_STOPPING_ERR_MSG Error message to finalize snapshot tasks.

Fields inherited from class org.apache.ignite.internal.processors.cache.GridCacheSharedManagerAdapter

cctx, diagnosticLog, log

Constructor Summary

Constructors

Constructor and Description
IgniteSnapshotManager(GridKernalContext ctx)

Method Summary

Modifier and Type	Method and Description
Set<UUID>	cacheStartRequiredAliveNodes(@Nullable IgniteUuid restoreId)
IgniteFuture<Boolean>	cancelLocalRestoreTask(String name)
void	cancelLocalSnapshotTask(String name)
IgniteFuture<Void>	cancelSnapshot(String name) Cancel running snapshot operation.
IgniteFuture<Boolean>	cancelSnapshotRestore(String name) Cancel snapshot restore operation.
IgniteInternalFuture<IdleVerifyResultV2>	checkSnapshot(String name, @Nullable String snpPath)
IgniteInternalFuture<SnapshotPartitionsVerifyTaskResult>	checkSnapshot(String name, @Nullable String snpPath, @Nullable Collection<String> grps, boolean includeCustomHandlers) The check snapshot procedure performs compute operation over the whole cluster to verify the snapshot entirety and partitions consistency.
IgniteFuture<Void>	createSnapshot(String name) Create a consistent copy of all persistence cache groups from the whole cluster.
IgniteFuture<Void>	createSnapshot(String name, @Nullable String snpPath) Create a consistent copy of all persistence cache groups from the whole cluster.
StandaloneGridKernalContext	createStandaloneKernalContext(File snpDir, String folderName)
@Nullable SnapshotOperationRequest	currentCreateRequest()
void	deleteSnapshot(File snpDir, String folderName)
protected IgniteSnapshotManager.SnapshotHandlers	handlers()
FileIOFactory	ioFactory()
void	ioFactory(FileIOFactory ioFactory)
boolean	isRestoring() Check if snapshot restore process is currently running.
boolean	isRestoring(CacheConfiguration<?,?> ccfg) Check if the cache or group with the specified name is currently being restored from the snapshot.
boolean	isRestoring(String snpName) Check if snapshot restore process is currently running.
boolean	isSnapshotCreating()
static boolean	isSnapshotOperation(DiscoveryEvent evt)
List<String>	localSnapshotNames()
void	onActivate(GridKernalContext kctx) Called when cluster performing activation.
void	onCacheGroupsStopped(List<Integer> grps)
void	onDeActivate(GridKernalContext kctx) Called when cluster performing deactivation.
void	onDoneBeforeTopologyUnlock(GridDhtPartitionsExchangeFuture fut) Callback from exchange process completion; called before topology is unlocked.
void	onReadyForRead(ReadOnlyMetastorage metaStorage) Is called when metastorage is made ready for read-only operations very early on node startup phase.
void	onReadyForReadWrite(ReadWriteMetastorage metaStorage) Fully functional metastore capable of performing reading and writing operations.
static File	partDeltaFile(File snapshotCacheDir, int partId)
static String	partDeltaFileName(int partId)
GridCloseableIterator<CacheDataRow>	partitionRowIterator(GridKernalContext ctx, String grpName, int partId, FilePageStore pageStore)
GridCloseableIterator<CacheDataRow>	partitionRowIterator(String snpName, String folderName, String grpName, int partId, @Nullable EncryptionCacheKeyProvider encrKeyProvider)
SnapshotMetadata	readSnapshotMetadata(File snpDir, String consId)
List<SnapshotMetadata>	readSnapshotMetadatas(String snpName, @Nullable String snpPath)
IgniteInternalFuture<Void>	requestRemoteSnapshotFiles(UUID rmtNodeId, String snpName, @Nullable String rmtSnpPath, Map<Integer,Set<Integer>> parts, BooleanSupplier stopChecker, BiConsumer<File,Throwable> partHnd)
static File	resolveSnapshotWorkDirectory(IgniteConfiguration cfg)
IgniteFuture<Void>	restoreSnapshot(String name, @Nullable Collection<String> grpNames) Restore cache group(s) from the snapshot.
IgniteFuture<Void>	restoreSnapshot(String name, @Nullable String snpPath, @Nullable Collection<String> grpNames) Restore cache group(s) from the snapshot.
IgniteFuture<Boolean>	restoreStatus(String snpName) Status of the restore operation cluster-wide.
List<File>	snapshotCacheDirectories(String snpName, @Nullable String snpPath, String folderName, Predicate<String> names)
File	snapshotLocalDir(String snpName)
File	snapshotLocalDir(String snpName, @Nullable String snpPath)
File	snapshotTmpDir()
protected void	start0()
protected void	stop0(boolean cancel)

Methods inherited from class org.apache.ignite.internal.processors.cache.GridCacheSharedManagerAdapter

context, isStopping, kernalStartInfo, kernalStopInfo, log, onDisconnected, onKernalStart, onKernalStart0, onKernalStop, onKernalStop0, onReconnected, printMemoryStats, start, startInfo, stop, stopInfo, toString

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.apache.ignite.internal.processors.cache.distributed.dht.preloader.PartitionsExchangeAware

onDoneAfterTopologyUnlock, onInitAfterTopologyLock, onInitBeforeTopologyLock

Field Detail

DELTA_SUFFIX

public static final String DELTA_SUFFIX

File with delta pages suffix.

See Also:

Constant Field Values

PART_DELTA_TEMPLATE

public static final String PART_DELTA_TEMPLATE

File name template consists of delta pages.

See Also:

Constant Field Values

INDEX_DELTA_NAME

public static final String INDEX_DELTA_NAME

File name template for index delta pages.

See Also:

Constant Field Values

CP_SNAPSHOT_REASON

public static final String CP_SNAPSHOT_REASON

Text Reason for checkpoint to start snapshot operation.

See Also:

Constant Field Values

RMT_SNAPSHOT_PREFIX

public static final String RMT_SNAPSHOT_PREFIX

Name prefix for each remote snapshot operation.

See Also:

Constant Field Values

DFLT_SNAPSHOT_TMP_DIR

public static final String DFLT_SNAPSHOT_TMP_DIR

Default snapshot directory for loading remote snapshots.

See Also:

Constant Field Values

SNP_IN_PROGRESS_ERR_MSG

public static final String SNP_IN_PROGRESS_ERR_MSG

Snapshot in progress error message.

See Also:

Constant Field Values

SNP_NODE_STOPPING_ERR_MSG

public static final String SNP_NODE_STOPPING_ERR_MSG

Error message to finalize snapshot tasks.

See Also:

Constant Field Values

SNAPSHOT_METRICS

public static final String SNAPSHOT_METRICS

Snapshot metrics prefix.

See Also:

Constant Field Values

SNAPSHOT_METAFILE_EXT

public static final String SNAPSHOT_METAFILE_EXT

Snapshot metafile extension.

See Also:

Constant Field Values

SNAPSHOT_RUNNER_THREAD_PREFIX

public static final String SNAPSHOT_RUNNER_THREAD_PREFIX

Prefix for snapshot threads.

See Also:

Constant Field Values

SNAPSHOT_TRANSFER_RATE_DMS_KEY

public static final String SNAPSHOT_TRANSFER_RATE_DMS_KEY

Snapshot transfer rate distributed configuration key

See Also:

Constant Field Values

DFLT_SNAPSHOT_TRANSFER_RATE_BYTES

public static final long DFLT_SNAPSHOT_TRANSFER_RATE_BYTES

Snapshot transfer rate is unlimited by default.

See Also:

Constant Field Values

SNAPSHOT_LIMITED_TRANSFER_BLOCK_SIZE_BYTES

public static final int SNAPSHOT_LIMITED_TRANSFER_BLOCK_SIZE_BYTES

Maximum block size for limited snapshot transfer (64KB by default).

See Also:

Constant Field Values

Constructor Detail

IgniteSnapshotManager

public IgniteSnapshotManager(GridKernalContext ctx)

Parameters:

ctx - Kernal context.

Method Detail

partDeltaFile

public static File partDeltaFile(File snapshotCacheDir,
                                 int partId)

Parameters:

snapshotCacheDir - Snapshot directory to store files.

partId - Cache partition identifier.

Returns:

A file representation.

partDeltaFileName

public static String partDeltaFileName(int partId)

Parameters:

partId - Partition id.

Returns:

File name of delta partition pages.

start0

protected void start0()
               throws IgniteCheckedException

Overrides:

start0 in class GridCacheSharedManagerAdapter

Throws:

IgniteCheckedException - If failed.

stop0

protected void stop0(boolean cancel)

Overrides:

stop0 in class GridCacheSharedManagerAdapter

Parameters:

cancel - Cancel flag.

onActivate

public void onActivate(GridKernalContext kctx)

Called when cluster performing activation.

Specified by:

onActivate in interface IgniteChangeGlobalStateSupport

Parameters:

kctx - Kernal context.

onDeActivate

public void onDeActivate(GridKernalContext kctx)

Called when cluster performing deactivation.

Specified by:

onDeActivate in interface IgniteChangeGlobalStateSupport

Parameters:

kctx - Kernal context.

deleteSnapshot

public void deleteSnapshot(File snpDir,
                           String folderName)

Parameters:

snpDir - Snapshot dir.

folderName - Local node folder name (see IgniteUtils.maskForFileName(java.lang.CharSequence) with consistent id).

snapshotLocalDir

public File snapshotLocalDir(String snpName)

Parameters:

snpName - Snapshot name.

Returns:

Local snapshot directory for snapshot with given name.

snapshotLocalDir

public File snapshotLocalDir(String snpName,
                             @Nullable
                             @Nullable String snpPath)

Parameters:

snpName - Snapshot name.

snpPath - Snapshot directory path.

Returns:

Local snapshot directory where snapshot files are located.

snapshotTmpDir

public File snapshotTmpDir()

Returns:

Node snapshot working directory.

isSnapshotCreating

public boolean isSnapshotCreating()

Returns:

True if snapshot operation is in progress.

currentCreateRequest

@Nullable
public @Nullable SnapshotOperationRequest currentCreateRequest()

Returns:

Current create snapshot request. Null if there is no create snapshot operation in progress.

isRestoring

public boolean isRestoring()

Check if snapshot restore process is currently running.

Returns:

True if the snapshot restore operation is in progress.

isRestoring

public boolean isRestoring(String snpName)

Check if snapshot restore process is currently running.

Parameters:

snpName - Snapshot name.

Returns:

True if the snapshot restore operation from the specified snapshot is in progress locally.

isRestoring

public boolean isRestoring(CacheConfiguration<?,?> ccfg)

Check if the cache or group with the specified name is currently being restored from the snapshot.

Parameters:

ccfg - Cache configuration.

Returns:

True if the cache or group with the specified name is being restored.

restoreStatus

public IgniteFuture<Boolean> restoreStatus(String snpName)

Status of the restore operation cluster-wide.

Parameters:

snpName - Snapshot name.

Returns:

Future that will be completed when the status of the restore operation is received from all the server nodes. The result of this future will be false if the restore process with the specified snapshot name is not running on all nodes.

cacheStartRequiredAliveNodes

public Set<UUID> cacheStartRequiredAliveNodes(@Nullable
                                              @Nullable IgniteUuid restoreId)

Parameters:

restoreId - Restore process ID.

Returns:

Server nodes on which a successful start of the cache(s) is required, if any of these nodes fails when starting the cache(s), the whole procedure is rolled back.

localSnapshotNames

public List<String> localSnapshotNames()

Returns:

List of all known snapshots on the local node.

cancelSnapshot

public IgniteFuture<Void> cancelSnapshot(String name)

Cancel running snapshot operation. All intermediate results of cancelled snapshot operation will be deleted. If snapshot already created this command will have no effect.

Specified by:

cancelSnapshot in interface IgniteSnapshot

Parameters:

name - Snapshot name to cancel.

Returns:

Future which will be completed when cancel operation finished.

cancelLocalSnapshotTask

public void cancelLocalSnapshotTask(String name)

Parameters:

name - Snapshot name to cancel operation on local node.

cancelSnapshotRestore

public IgniteFuture<Boolean> cancelSnapshotRestore(String name)

Cancel snapshot restore operation.

Specified by:

cancelSnapshotRestore in interface IgniteSnapshot

Parameters:

name - Snapshot name.

Returns:

Future that will be finished when the process is complete on all nodes. The result of this future will be false if the restore process with the specified snapshot name is not running at all.

cancelLocalRestoreTask

public IgniteFuture<Boolean> cancelLocalRestoreTask(String name)

Parameters:

name - Snapshot name.

Returns:

Future that will be finished when process the process is complete. The result of this future will be false if the restore process with the specified snapshot name is not running at all.

checkSnapshot

public IgniteInternalFuture<IdleVerifyResultV2> checkSnapshot(String name,
                                                              @Nullable
                                                              @Nullable String snpPath)

Parameters:

name - Snapshot name.

snpPath - Snapshot directory path.

Returns:

Future with the result of execution snapshot partitions verify task, which besides calculating partition hashes of IdleVerifyResultV2 also contains the snapshot metadata distribution across the cluster.

checkSnapshot

public IgniteInternalFuture<SnapshotPartitionsVerifyTaskResult> checkSnapshot(String name,
                                                                              @Nullable
                                                                              @Nullable String snpPath,
                                                                              @Nullable
                                                                              @Nullable Collection<String> grps,
                                                                              boolean includeCustomHandlers)

The check snapshot procedure performs compute operation over the whole cluster to verify the snapshot entirety and partitions consistency. The result future will be completed with an exception if this exception is not related to the check procedure, and will be completed normally with the IdleVerifyResult.

Parameters:

name - Snapshot name.

snpPath - Snapshot directory path.

grps - Collection of cache group names to check.

includeCustomHandlers - True to invoke all user-defined SnapshotHandlerType.RESTORE handlers, otherwise only system consistency check will be performed.

Returns:

Future with the result of execution snapshot partitions verify task, which besides calculating partition hashes of IdleVerifyResultV2 also contains the snapshot metadata distribution across the cluster.

snapshotCacheDirectories

public List<File> snapshotCacheDirectories(String snpName,
                                           @Nullable
                                           @Nullable String snpPath,
                                           String folderName,
                                           Predicate<String> names)

Parameters:

snpName - Snapshot name.

folderName - The name of a directory for the cache group.

names - Cache group names to filter.

Returns:

The list of cache or cache group names in given snapshot on local node.

readSnapshotMetadata

public SnapshotMetadata readSnapshotMetadata(File snpDir,
                                             String consId)

Parameters:

snpDir - The full path to the snapshot files.

consId - Node consistent id to read metadata for.

Returns:

Snapshot metadata instance.

readSnapshotMetadatas

public List<SnapshotMetadata> readSnapshotMetadatas(String snpName,
                                                    @Nullable
                                                    @Nullable String snpPath)

Parameters:

snpName - Snapshot name.

snpPath - Snapshot directory path.

Returns:

List of snapshot metadata for the given snapshot name on local node. If snapshot has been taken from local node the snapshot metadata for given local node will be placed on the first place.

createSnapshot

public IgniteFuture<Void> createSnapshot(String name)

Create a consistent copy of all persistence cache groups from the whole cluster.

Specified by:

createSnapshot in interface IgniteSnapshot

Parameters:

name - Snapshot unique name which satisfies the following name pattern [a-zA-Z0-9_].

Returns:

Future which will be completed when a process ends.

createSnapshot

public IgniteFuture<Void> createSnapshot(String name,
                                         @Nullable
                                         @Nullable String snpPath)

Create a consistent copy of all persistence cache groups from the whole cluster.

Parameters:

name - Snapshot unique name which satisfies the following name pattern [a-zA-Z0-9_].

snpPath - Snapshot directory path.

Returns:

Future which will be completed when a process ends.

restoreSnapshot

public IgniteFuture<Void> restoreSnapshot(String name,
                                          @Nullable
                                          @Nullable Collection<String> grpNames)

Restore cache group(s) from the snapshot.

NOTE: Cache groups to be restored from the snapshot must not present in the cluster, if they present, they must be destroyed by the user (eg with IgniteCache.destroy()) before starting this operation.

Specified by:

restoreSnapshot in interface IgniteSnapshot

Parameters:

name - Snapshot name.

grpNames - Cache groups to be restored or null to restore all cache groups from the snapshot.

Returns:

Future which will be completed when restore operation finished.

restoreSnapshot

public IgniteFuture<Void> restoreSnapshot(String name,
                                          @Nullable
                                          @Nullable String snpPath,
                                          @Nullable
                                          @Nullable Collection<String> grpNames)

Restore cache group(s) from the snapshot.

Parameters:

name - Snapshot name.

snpPath - Snapshot directory path.

grpNames - Cache groups to be restored or null to restore all cache groups from the snapshot.

Returns:

Future which will be completed when restore operation finished.

onReadyForReadWrite

public void onReadyForReadWrite(ReadWriteMetastorage metaStorage)
                         throws IgniteCheckedException

Fully functional metastore capable of performing reading and writing operations. Components interested in using metastore are allowed to keep reference passed into the method in their fields.

Specified by:

onReadyForReadWrite in interface MetastorageLifecycleListener

Parameters:

metaStorage - Fully functional meta storage.

Throws:

IgniteCheckedException

onReadyForRead

public void onReadyForRead(ReadOnlyMetastorage metaStorage)
                    throws IgniteCheckedException

Is called when metastorage is made ready for read-only operations very early on node startup phase. Reference for read-only metastorage should be used only within this method and shouldn't be stored to any field.

Specified by:

onReadyForRead in interface MetastorageLifecycleListener

Parameters:

metaStorage - Read-only meta storage.

Throws:

IgniteCheckedException

isSnapshotOperation

public static boolean isSnapshotOperation(DiscoveryEvent evt)

Parameters:

evt - Discovery event to check.

Returns:

true if exchange started by snapshot operation.

onDoneBeforeTopologyUnlock

public void onDoneBeforeTopologyUnlock(GridDhtPartitionsExchangeFuture fut)

Callback from exchange process completion; called before topology is unlocked. Guarantees that no updates were performed on local node since exchange process started.

Specified by:

onDoneBeforeTopologyUnlock in interface PartitionsExchangeAware

Parameters:

fut - Partition map exchange future.

requestRemoteSnapshotFiles

public IgniteInternalFuture<Void> requestRemoteSnapshotFiles(UUID rmtNodeId,
                                                             String snpName,
                                                             @Nullable
                                                             @Nullable String rmtSnpPath,
                                                             Map<Integer,Set<Integer>> parts,
                                                             BooleanSupplier stopChecker,
                                                             BiConsumer<File,Throwable> partHnd)
                                                      throws IgniteCheckedException

Parameters:

rmtNodeId - The remote node to connect to.

snpName - Snapshot name to request.

rmtSnpPath - Snapshot directory path on the remote node.

parts - Collection of pairs group and appropriate cache partition to be snapshot.

stopChecker - Node stop or prcoess interrupt checker.

partHnd - Received partition handler.

Throws:

IgniteCheckedException

onCacheGroupsStopped

public void onCacheGroupsStopped(List<Integer> grps)

Parameters:

grps - List of cache groups which will be destroyed.

createStandaloneKernalContext

public StandaloneGridKernalContext createStandaloneKernalContext(File snpDir,
                                                                 String folderName)
                                                          throws IgniteCheckedException

Parameters:

snpDir - The full path to the snapshot files.

folderName - The node folder name, usually it's the same as the U.maskForFileName(consistentId).

Returns:

Standalone kernal context related to the snapshot.

Throws:

IgniteCheckedException - If fails.

partitionRowIterator

public GridCloseableIterator<CacheDataRow> partitionRowIterator(GridKernalContext ctx,
                                                                String grpName,
                                                                int partId,
                                                                FilePageStore pageStore)
                                                         throws IgniteCheckedException

Parameters:

grpName - Cache group name.

partId - Partition id.

pageStore - File page store to iterate over.

Returns:

Iterator over partition.

Throws:

IgniteCheckedException - If and error occurs.

partitionRowIterator

public GridCloseableIterator<CacheDataRow> partitionRowIterator(String snpName,
                                                                String folderName,
                                                                String grpName,
                                                                int partId,
                                                                @Nullable
                                                                @Nullable EncryptionCacheKeyProvider encrKeyProvider)
                                                         throws IgniteCheckedException

Parameters:

snpName - Snapshot name.

folderName - The node folder name, usually it's the same as the U.maskForFileName(consistentId).

grpName - Cache group name.

partId - Partition id.

encrKeyProvider - Encryption keys provider to create encrypted IO. If null, no encrypted IO is used.

Returns:

Iterator over partition.

Throws:

IgniteCheckedException - If and error occurs.

ioFactory

public void ioFactory(FileIOFactory ioFactory)

Parameters:

ioFactory - Factory to create IO interface over a page stores.

ioFactory

public FileIOFactory ioFactory()

Returns:

Factory to create IO interface over a page stores.

resolveSnapshotWorkDirectory

public static File resolveSnapshotWorkDirectory(IgniteConfiguration cfg)

Parameters:

cfg - Ignite configuration.

Returns:

Snapshot directory resolved through given configuration.

handlers

protected IgniteSnapshotManager.SnapshotHandlers handlers()

Returns:

Snapshot handlers.