io.aeron.cluster.RecordingLog Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of aeron-agent Show documentation
Show all versions of aeron-agent Show documentation
Efficient reliable UDP unicast, UDP multicast, and IPC transport protocol.
/*
* Copyright 2014-2023 Real Logic Limited.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* https://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package io.aeron.cluster;
import io.aeron.Aeron;
import io.aeron.archive.client.AeronArchive;
import io.aeron.archive.status.RecordingPos;
import io.aeron.cluster.client.ClusterException;
import org.agrona.BitUtil;
import org.agrona.CloseHelper;
import org.agrona.DirectBuffer;
import org.agrona.LangUtil;
import org.agrona.Strings;
import org.agrona.collections.IntArrayList;
import org.agrona.collections.Long2LongHashMap;
import org.agrona.collections.MutableReference;
import org.agrona.collections.Object2ObjectHashMap;
import org.agrona.concurrent.UnsafeBuffer;
import java.io.File;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.nio.channels.FileChannel;
import java.nio.file.StandardOpenOption;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.EnumSet;
import java.util.List;
import java.util.Map;
import java.util.NavigableMap;
import java.util.Objects;
import java.util.TreeMap;
import static io.aeron.Aeron.NULL_VALUE;
import static io.aeron.archive.client.AeronArchive.NULL_POSITION;
import static io.aeron.driver.Configuration.FILE_PAGE_SIZE_DEFAULT;
import static java.lang.Math.max;
import static java.nio.ByteOrder.LITTLE_ENDIAN;
import static java.nio.file.StandardOpenOption.*;
import static org.agrona.BitUtil.*;
/**
* A log of recordings which make up the history of a Raft log across leadership terms. Entries are in order.
*
* The log is made up of entries of leadership terms or snapshots to roll up state as of a log position within a
* leadership term.
*
* The latest state is made up of the latest snapshot followed by any leadership log which follows. It is possible
* that a snapshot is taken midterm and therefore the latest state is the snapshot plus the log of messages which
* got appended to the log after the snapshot was taken.
*
* Record layout as follows:
*
* 0 1 2 3
* 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
* | Recording ID |
* | |
* +---------------------------------------------------------------+
* | Leadership Term ID |
* | |
* +---------------------------------------------------------------+
* | Log Position at beginning of term |
* | |
* +---------------------------------------------------------------+
* | Log Position when entry was created or committed |
* | |
* +---------------------------------------------------------------+
* | Timestamp when entry was created |
* | |
* +---------------------------------------------------------------+
* | Service ID when a Snapshot |
* +---------------------------------------------------------------+
* |R| Entry Type (Log or Snapshot) |
* +---------------------------------------------------------------+
* | Archive Endpoint (length) |
* +---------------------------------------------------------------+
* | Archive Endpoint (length) |
* | Archive Endpoint ...
* ... (variable) |
* +---------------------------------------------------------------+
* | |
* | ...
* ... Repeats to the end of the log |
* | |
* +---------------------------------------------------------------+
*
* The reserved bit on the entry type indicates whether the entry was marked invalid.
*/
public final class RecordingLog implements AutoCloseable
{
/**
* Representation of the entry in the {@link RecordingLog}.
*/
public static final class Entry
{
/**
* Identity of the recording in the archive.
*/
public final long recordingId;
/**
* Identity of the leadership term.
*/
public final long leadershipTermId;
/**
* The log position at the base of the leadership term.
*/
public final long termBaseLogPosition;
/**
* Position the log has reached for the entry.
*/
public final long logPosition;
/**
* Timestamp for the cluster clock in the time units configured for the cluster.
*/
public final long timestamp;
/**
* Identity of the service associated with the entry.
*/
public final int serviceId;
/**
* Type, or classification, of the entry, e.g. {@link #ENTRY_TYPE_TERM} or {@link #ENTRY_TYPE_SNAPSHOT}.
*/
public final int type;
/**
* Index of the entry in the recording log.
*/
public final int entryIndex;
/**
* Flag to indicate if the entry is invalid and thus should be ignored.
*/
public final boolean isValid;
/**
* Endpoint for an archive where a remote snapshot is located.
*/
public final String archiveEndpoint;
private long position;
/**
* A new entry in the recording log.
*
* @param recordingId of the entry in an archive.
* @param leadershipTermId of this entry.
* @param termBaseLogPosition position of the log over leadership terms at the beginning of this term.
* @param logPosition position reached when the entry was created
* @param timestamp of this entry.
* @param serviceId service id for snapshot.
* @param type of the entry as a log of a term or a snapshot.
* @param archiveEndpoint archive where the snapshot is located, if
* entryType == ENTRY_TYPE_STANDBY_SNAPSHOT.
* @param isValid indicates if the entry is valid, {@link RecordingLog#invalidateEntry(long, int)}
* marks it invalid.
* @param position of the entry on disk.
* @param entryIndex of the entry on disk.
* @throws ClusterException if entryType == ENTRY_TYPE_STANDBY_SNAPSHOT and endpoint
* is null or empty.
*/
public Entry(
final long recordingId,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long timestamp,
final int serviceId,
final int type,
final String archiveEndpoint,
final boolean isValid,
final long position,
final int entryIndex)
{
this.recordingId = recordingId;
this.leadershipTermId = leadershipTermId;
this.termBaseLogPosition = termBaseLogPosition;
this.logPosition = logPosition;
this.timestamp = timestamp;
this.serviceId = serviceId;
this.type = type;
this.archiveEndpoint = archiveEndpoint;
this.position = position;
this.entryIndex = entryIndex;
this.isValid = isValid;
if (ENTRY_TYPE_STANDBY_SNAPSHOT == type && Strings.isEmpty(archiveEndpoint))
{
throw new ClusterException("Remote snapshots must has a valid endpoint");
}
}
Entry(
final long recordingId,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long timestamp,
final int serviceId,
final int type,
final String archiveEndpoint,
final boolean isValid,
final int entryIndex)
{
this(
recordingId,
leadershipTermId,
termBaseLogPosition,
logPosition,
timestamp,
serviceId,
type,
archiveEndpoint,
isValid,
NULL_VALUE,
entryIndex);
}
/**
* Binary length of the serialised entry including alignment.
*
* @return length of this entry.
*/
public int length()
{
final int unalignedLength = (ENTRY_TYPE_STANDBY_SNAPSHOT == type) ?
ENDPOINT_OFFSET + SIZE_OF_INT + archiveEndpoint.length() : ENDPOINT_OFFSET;
return align(unalignedLength, CACHE_LINE_LENGTH);
}
private void position(final long position)
{
this.position = position;
}
Entry invalidate()
{
return new Entry(
recordingId,
leadershipTermId,
termBaseLogPosition,
logPosition,
timestamp,
serviceId,
type,
archiveEndpoint,
false,
position,
entryIndex);
}
Entry logPosition(final long logPosition)
{
return new Entry(
recordingId,
leadershipTermId,
termBaseLogPosition,
logPosition,
timestamp,
serviceId,
type,
archiveEndpoint,
isValid,
position,
entryIndex);
}
long serviceId()
{
return serviceId;
}
/**
* {@inheritDoc}
*/
public boolean equals(final Object o)
{
if (this == o)
{
return true;
}
if (o == null || getClass() != o.getClass())
{
return false;
}
final Entry entry = (Entry)o;
return recordingId == entry.recordingId &&
leadershipTermId == entry.leadershipTermId &&
termBaseLogPosition == entry.termBaseLogPosition &&
logPosition == entry.logPosition &&
timestamp == entry.timestamp &&
serviceId == entry.serviceId &&
type == entry.type &&
entryIndex == entry.entryIndex &&
isValid == entry.isValid;
}
/**
* {@inheritDoc}
*/
public int hashCode()
{
int result = (int)(recordingId ^ (recordingId >>> 32));
result = 31 * result + (int)(leadershipTermId ^ (leadershipTermId >>> 32));
result = 31 * result + (int)(termBaseLogPosition ^ (termBaseLogPosition >>> 32));
result = 31 * result + (int)(logPosition ^ (logPosition >>> 32));
result = 31 * result + (int)(timestamp ^ (timestamp >>> 32));
result = 31 * result + serviceId;
result = 31 * result + type;
result = 31 * result + entryIndex;
result = 31 * result + (isValid ? 1 : 0);
return result;
}
/**
* {@inheritDoc}
*/
public String toString()
{
final String archiveEndpointEntry = ENTRY_TYPE_STANDBY_SNAPSHOT == type ?
", archiveEndpoint=" + archiveEndpoint : "";
return "Entry{" +
"recordingId=" + recordingId +
", leadershipTermId=" + leadershipTermId +
", termBaseLogPosition=" + termBaseLogPosition +
", logPosition=" + logPosition +
", timestamp=" + timestamp +
", serviceId=" + serviceId +
", type=" + typeAsString(type) +
", isValid=" + isValid +
", entryIndex=" + entryIndex +
archiveEndpointEntry +
'}';
}
}
/**
* Representation of a snapshot entry in the {@link RecordingLog}.
*/
public static final class Snapshot
{
/**
* Identity of the recording in the archive for the snapshot.
*/
public final long recordingId;
/**
* Identity of the leadership term.
*/
public final long leadershipTermId;
/**
* The log position at the base of the leadership term.
*/
public final long termBaseLogPosition;
/**
* Position the log has reached for the snapshot.
*/
public final long logPosition;
/**
* Timestamp for the cluster clock in the time units configured for the cluster at time of the snapshot.
*/
public final long timestamp;
/**
* Identity of the service associated with the snapshot.
*/
public final int serviceId;
/**
* A snapshot entry in the {@link RecordingLog}.
*
* @param recordingId of the entry in an archive.
* @param leadershipTermId in which the snapshot was taken.
* @param termBaseLogPosition position of the log over leadership terms at the beginning of this term.
* @param logPosition position reached when the entry was snapshot was taken.
* @param timestamp as which the snapshot was taken.
* @param serviceId which the snapshot belongs to.
*/
public Snapshot(
final long recordingId,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long timestamp,
final int serviceId)
{
this.recordingId = recordingId;
this.leadershipTermId = leadershipTermId;
this.termBaseLogPosition = termBaseLogPosition;
this.logPosition = logPosition;
this.timestamp = timestamp;
this.serviceId = serviceId;
}
/**
* {@inheritDoc}
*/
public String toString()
{
return "Snapshot{" +
"recordingId=" + recordingId +
", leadershipTermId=" + leadershipTermId +
", termBaseLogPosition=" + termBaseLogPosition +
", logPosition=" + logPosition +
", timestamp=" + timestamp +
", serviceId=" + serviceId +
'}';
}
}
/**
* Representation of a log entry in the {@link RecordingLog}.
*/
public static final class Log
{
/**
* Identity of the recording in the archive for the log.
*/
public final long recordingId;
/**
* Identity of the leadership term.
*/
public final long leadershipTermId;
/**
* The log position at the base of the leadership term.
*/
public final long termBaseLogPosition;
/**
* Position the log has reached for the term which can be {@link AeronArchive#NULL_POSITION} when not committed.
*/
public final long logPosition;
/**
* Start position of the recording captured in the archive.
*/
public final long startPosition;
/**
* Stop position of the recording captured in the archive.
*/
public final long stopPosition;
/**
* Initial term identity of the stream captured for the recording.
*/
public final int initialTermId;
/**
* Transport term buffer length of the stream captured for the recording.
*/
public final int termBufferLength;
/**
* Transport MTU length of the stream captured for the recording.
*/
public final int mtuLength;
/**
* Transport session identity of the stream captured for the recording.
*/
public final int sessionId;
/**
* Construct a representation of a log entry in the {@link RecordingLog}.
*
* @param recordingId for the recording in an archive.
* @param leadershipTermId identity for the leadership term.
* @param termBaseLogPosition log position at the base of the leadership term.
* @param logPosition position the log has reached for the term.
* @param startPosition of the recording captured in the archive.
* @param stopPosition of the recording captured in the archive.
* @param initialTermId of the stream captured for the recording.
* @param termBufferLength of the stream captured for the recording.
* @param mtuLength of the stream captured for the recording.
* @param sessionId of the stream captured for the recording.
*/
public Log(
final long recordingId,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long startPosition,
final long stopPosition,
final int initialTermId,
final int termBufferLength,
final int mtuLength,
final int sessionId)
{
this.recordingId = recordingId;
this.leadershipTermId = leadershipTermId;
this.termBaseLogPosition = termBaseLogPosition;
this.logPosition = logPosition;
this.startPosition = startPosition;
this.stopPosition = stopPosition;
this.initialTermId = initialTermId;
this.termBufferLength = termBufferLength;
this.mtuLength = mtuLength;
this.sessionId = sessionId;
}
/**
* {@inheritDoc}
*/
public String toString()
{
return "Log{" +
"recordingId=" + recordingId +
", leadershipTermId=" + leadershipTermId +
", termBaseLogPosition=" + termBaseLogPosition +
", logPosition=" + logPosition +
", startPosition=" + startPosition +
", stopPosition=" + stopPosition +
", initialTermId=" + initialTermId +
", termBufferLength=" + termBufferLength +
", mtuLength=" + mtuLength +
", sessionId=" + sessionId +
'}';
}
}
/**
* The snapshots and steps to recover the state of a cluster.
*/
public static final class RecoveryPlan
{
/**
* The last, i.e. most recent, leadership term identity for the log.
*/
public final long lastLeadershipTermId;
/**
* The last, i.e. most recent, leadership term base log position.
*/
public final long lastTermBaseLogPosition;
/**
* The position reached for local appended log.
*/
public final long appendedLogPosition;
/**
* The position reached for the local appended log for which the commit position is known.
*/
public final long committedLogPosition;
/**
* The most recent snapshots for the consensus module and services to accelerate recovery.
*/
public final ArrayList snapshots;
/**
* The appended local log details.
*/
public final Log log;
/**
* @param lastLeadershipTermId the last, i.e. most recent, leadership term identity for the log.
* @param lastTermBaseLogPosition last, i.e. most recent, leadership term base log position.
* @param appendedLogPosition reached for local appended log.
* @param committedLogPosition reached for the local appended log for which the commit position is known.
* @param snapshots most recent snapshots for the consensus module and services.
* @param log appended local log details.
*/
public RecoveryPlan(
final long lastLeadershipTermId,
final long lastTermBaseLogPosition,
final long appendedLogPosition,
final long committedLogPosition,
final ArrayList snapshots,
final Log log)
{
this.lastLeadershipTermId = lastLeadershipTermId;
this.lastTermBaseLogPosition = lastTermBaseLogPosition;
this.appendedLogPosition = appendedLogPosition;
this.committedLogPosition = committedLogPosition;
this.snapshots = snapshots;
this.log = log;
}
/**
* {@inheritDoc}
*/
public String toString()
{
return "RecoveryPlan{" +
"lastLeadershipTermId=" + lastLeadershipTermId +
", lastTermBaseLogPosition=" + lastTermBaseLogPosition +
", appendedLogPosition=" + appendedLogPosition +
", committedLogPosition=" + committedLogPosition +
", snapshots=" + snapshots +
", log=" + log +
'}';
}
}
/**
* Filename for the history of leadership log terms and snapshot recordings.
*/
public static final String RECORDING_LOG_FILE_NAME = "recording.log";
/**
* The log entry is for a recording of messages within a leadership term to the log.
*/
public static final int ENTRY_TYPE_TERM = 0;
/**
* The log entry is for a recording of a snapshot of state taken as of a position in the log.
*/
public static final int ENTRY_TYPE_SNAPSHOT = 1;
/**
* The log entry is for a recording of a snapshot of state taken as of a position in the log on another machine.
* Entries of this time should have an endpoint for an archive associated with them.
*/
public static final int ENTRY_TYPE_STANDBY_SNAPSHOT = 2;
/**
* The flag used to determine if the entry has been marked with invalid.
*/
public static final int ENTRY_TYPE_INVALID_FLAG = 1 << 31;
/**
* The offset at which the recording id for the entry is stored.
*/
public static final int RECORDING_ID_OFFSET = 0;
/**
* The offset at which the leadership term id for the entry is stored.
*/
public static final int LEADERSHIP_TERM_ID_OFFSET = RECORDING_ID_OFFSET + SIZE_OF_LONG;
/**
* The offset at which the log position as of the beginning of the term for the entry is stored.
*/
public static final int TERM_BASE_LOG_POSITION_OFFSET = LEADERSHIP_TERM_ID_OFFSET + SIZE_OF_LONG;
/**
* The offset at which the log position is stored.
*/
public static final int LOG_POSITION_OFFSET = TERM_BASE_LOG_POSITION_OFFSET + SIZE_OF_LONG;
/**
* The offset at which the timestamp for the entry is stored.
*/
public static final int TIMESTAMP_OFFSET = LOG_POSITION_OFFSET + SIZE_OF_LONG;
/**
* The offset at which the service id is recorded.
*/
public static final int SERVICE_ID_OFFSET = TIMESTAMP_OFFSET + SIZE_OF_LONG;
/**
* The offset at which the type of the entry is stored.
*/
public static final int ENTRY_TYPE_OFFSET = SERVICE_ID_OFFSET + SIZE_OF_INT;
/**
* The offset at which the endpoint of the remote snapshot is held.
*/
public static final int ENDPOINT_OFFSET = ENTRY_TYPE_OFFSET + SIZE_OF_INT;
private static final int MAX_ENDPOINT_LENGTH = SIZE_OF_INT + 255 + 6;
/**
* Maximum possible entry length. Include the entry plus a variable length endpoint string.
*/
public static final int MAX_ENTRY_LENGTH = BitUtil.align(ENDPOINT_OFFSET + MAX_ENDPOINT_LENGTH, CACHE_LINE_LENGTH);
private static final Comparator ENTRY_COMPARATOR =
(Entry e1, Entry e2) ->
{
int result = Long.compare(e1.leadershipTermId, e2.leadershipTermId);
if (0 != result)
{
return result;
}
result = Integer.compare(e1.type, e2.type);
if (0 != result)
{
return result;
}
if (ENTRY_TYPE_TERM == e1.type)
{
return Integer.compare(e1.entryIndex, e2.entryIndex);
}
else
{
result = Long.compare(e1.logPosition, e2.logPosition);
if (0 != result)
{
return result;
}
return Integer.compare(e2.serviceId, e1.serviceId);
}
};
private long termRecordingId = NULL_VALUE;
private int nextEntryIndex;
private final FileChannel fileChannel;
private final ByteBuffer byteBuffer = ByteBuffer.allocateDirect(FILE_PAGE_SIZE_DEFAULT).order(LITTLE_ENDIAN);
private final UnsafeBuffer buffer = new UnsafeBuffer(byteBuffer);
private final ArrayList entriesCache = new ArrayList<>();
private final Long2LongHashMap cacheIndexByLeadershipTermIdMap = new Long2LongHashMap(NULL_VALUE);
private final IntArrayList invalidSnapshots = new IntArrayList();
/**
* Create a log that appends to an existing log or creates a new one.
*
* @param parentDir in which the log will be created.
* @param createNew create a new recording log if one does not exist.
*/
public RecordingLog(final File parentDir, final boolean createNew)
{
final File logFile = new File(parentDir, RECORDING_LOG_FILE_NAME);
final boolean isNewFile = !logFile.exists();
final EnumSet openOptions = EnumSet.of(READ, WRITE);
if (createNew)
{
openOptions.add(CREATE);
}
try
{
fileChannel = FileChannel.open(logFile.toPath(), openOptions);
if (isNewFile)
{
syncDirectory(parentDir);
}
else
{
reload();
}
}
catch (final IOException ex)
{
throw new ClusterException(ex);
}
}
/**
* {@inheritDoc}
*/
public void close()
{
CloseHelper.close(fileChannel);
}
/**
* Force the file to durable storage. Same as calling {@link FileChannel#force(boolean)} with true.
*
* @param fileSyncLevel as defined by {@link ConsensusModule.Configuration#FILE_SYNC_LEVEL_PROP_NAME}.
*/
public void force(final int fileSyncLevel)
{
if (fileSyncLevel > 0)
{
try
{
fileChannel.force(fileSyncLevel > 1);
}
catch (final IOException ex)
{
LangUtil.rethrowUnchecked(ex);
}
}
}
/**
* List of currently loaded entries.
*
* @return the list of currently loaded entries.
*/
public List entries()
{
return entriesCache;
}
/**
* Get the next index to be used when appending an entry to the log.
*
* @return the next index to be used when appending an entry to the log.
*/
public int nextEntryIndex()
{
return nextEntryIndex;
}
/**
* Reload the recording log from disk.
*/
public void reload()
{
entriesCache.clear();
cacheIndexByLeadershipTermIdMap.clear();
invalidSnapshots.clear();
cacheIndexByLeadershipTermIdMap.compact();
nextEntryIndex = 0;
byteBuffer.clear();
try
{
long filePosition = 0;
long consumePosition = 0;
while (true)
{
final int bytesRead = fileChannel.read(byteBuffer, filePosition);
filePosition += bytesRead;
if (0 < bytesRead)
{
byteBuffer.flip();
consumePosition += captureEntriesFromBuffer(consumePosition, byteBuffer, buffer, entriesCache);
byteBuffer.compact();
}
else
{
break;
}
}
}
catch (final IOException ex)
{
LangUtil.rethrowUnchecked(ex);
}
entriesCache.sort(ENTRY_COMPARATOR);
for (int i = 0, size = entriesCache.size(); i < size; i++)
{
final Entry entry = entriesCache.get(i);
if (isValidTerm(entry))
{
cacheIndexByLeadershipTermIdMap.put(entry.leadershipTermId, i);
}
if (isInvalidSnapshot(entry))
{
invalidSnapshots.add(i);
}
}
}
/**
* Find the last recording id used for a leadership term. If not found then {@link RecordingPos#NULL_RECORDING_ID}.
*
* @return the last leadership term recording id or {@link RecordingPos#NULL_RECORDING_ID} if not found.
*/
public long findLastTermRecordingId()
{
final Entry lastTerm = findLastTerm();
return null != lastTerm ? lastTerm.recordingId : RecordingPos.NULL_RECORDING_ID;
}
/**
* Find the last leadership term in the recording log.
*
* @return the last leadership term in the recording log.
*/
public Entry findLastTerm()
{
for (int i = entriesCache.size() - 1; i >= 0; i--)
{
final Entry entry = entriesCache.get(i);
if (isValidTerm(entry))
{
return entry;
}
}
return null;
}
/**
* Get the term {@link Entry} for a given leadership term id.
*
* @param leadershipTermId to get {@link Entry} for.
* @return the {@link Entry} if found or throw {@link IllegalArgumentException} if no entry exists for term.
*/
public Entry getTermEntry(final long leadershipTermId)
{
final int index = (int)cacheIndexByLeadershipTermIdMap.get(leadershipTermId);
if (NULL_VALUE != index)
{
return entriesCache.get(index);
}
throw new ClusterException("unknown leadershipTermId=" + leadershipTermId);
}
/**
* Find the term {@link Entry} for a given leadership term id.
*
* @param leadershipTermId to get {@link Entry} for.
* @return the {@link Entry} if found or null if not found.
*/
public Entry findTermEntry(final long leadershipTermId)
{
final int index = (int)cacheIndexByLeadershipTermIdMap.get(leadershipTermId);
if (NULL_VALUE != index)
{
return entriesCache.get(index);
}
return null;
}
/**
* Get the latest snapshot {@link Entry} in the log.
*
* @param serviceId for the snapshot.
* @return the latest snapshot {@link Entry} in the log or null if no snapshot exists.
*/
public Entry getLatestSnapshot(final int serviceId)
{
for (int i = entriesCache.size() - 1; i >= 0; i--)
{
final Entry entry = entriesCache.get(i);
if (isValidSnapshot(entry) && ConsensusModule.Configuration.SERVICE_ID == entry.serviceId)
{
if (ConsensusModule.Configuration.SERVICE_ID == serviceId)
{
return entry;
}
final int serviceSnapshotIndex = i - (serviceId + 1);
if (serviceSnapshotIndex >= 0)
{
final Entry snapshot = entriesCache.get(serviceSnapshotIndex);
if (isValidSnapshot(snapshot) && serviceId == snapshot.serviceId)
{
return snapshot;
}
}
}
}
return null;
}
/**
* Invalidate the last snapshot taken by the cluster so that on restart it can revert to the previous one.
*
* @return true if the latest snapshot was found and marked as invalid, so it will not be reloaded.
*/
public boolean invalidateLatestSnapshot()
{
int index = -1;
for (int i = entriesCache.size() - 1; i >= 0; i--)
{
final Entry entry = entriesCache.get(i);
if (isValidSnapshot(entry) && ConsensusModule.Configuration.SERVICE_ID == entry.serviceId)
{
if (!cacheIndexByLeadershipTermIdMap.containsKey(entry.leadershipTermId))
{
throw new ClusterException(
"no matching term for snapshot: leadershipTermId=" + entry.leadershipTermId);
}
index = i;
break;
}
}
if (index >= 0)
{
int serviceId = ConsensusModule.Configuration.SERVICE_ID;
for (int i = index; i >= 0; i--)
{
final Entry entry = entriesCache.get(i);
if (isValidSnapshot(entry) && entry.serviceId == serviceId)
{
invalidateEntry(entry.leadershipTermId, entry.entryIndex);
serviceId++;
}
else
{
break;
}
}
return true;
}
return false;
}
/**
* Get the {@link Entry#timestamp} for a term.
*
* @param leadershipTermId to get {@link Entry#timestamp} for.
* @return the timestamp or {@link io.aeron.Aeron#NULL_VALUE} if not found.
*/
public long getTermTimestamp(final long leadershipTermId)
{
final int index = (int)cacheIndexByLeadershipTermIdMap.get(leadershipTermId);
if (NULL_VALUE != index)
{
return entriesCache.get(index).timestamp;
}
return NULL_VALUE;
}
/**
* Create a recovery plan for the cluster so that when the steps are replayed the plan will bring the cluster
* back to the latest stable state.
*
* @param archive to lookup recording descriptors.
* @param serviceCount of services that may have snapshots.
* @param replicatedRecordingId leader's recordingId used for replicating
* @return a new {@link RecoveryPlan} for the cluster.
*/
public RecoveryPlan createRecoveryPlan(
final AeronArchive archive, final int serviceCount, final long replicatedRecordingId)
{
final ArrayList snapshots = new ArrayList<>();
final MutableReference logRef = new MutableReference<>();
planRecovery(snapshots, logRef, entriesCache, archive, serviceCount, replicatedRecordingId);
long lastLeadershipTermId = NULL_VALUE;
long lastTermBaseLogPosition = 0;
long committedLogPosition = 0;
long appendedLogPosition = 0;
final int snapshotStepsSize = snapshots.size();
if (snapshotStepsSize > 0)
{
final Snapshot snapshot = snapshots.get(0);
lastLeadershipTermId = snapshot.leadershipTermId;
lastTermBaseLogPosition = snapshot.termBaseLogPosition;
appendedLogPosition = snapshot.logPosition;
committedLogPosition = snapshot.logPosition;
}
final Log log = logRef.get();
if (null != log)
{
lastLeadershipTermId = log.leadershipTermId;
lastTermBaseLogPosition = log.termBaseLogPosition;
appendedLogPosition = log.stopPosition;
committedLogPosition = NULL_POSITION != log.logPosition ? log.logPosition : committedLogPosition;
}
return new RecoveryPlan(
lastLeadershipTermId,
lastTermBaseLogPosition,
appendedLogPosition,
committedLogPosition,
snapshots,
log);
}
/**
* Create a recovery plan that has only snapshots. Used for dynamicJoin snapshot load.
*
* @param snapshots to construct plan from.
* @return a new {@link RecoveryPlan} for the cluster.
*/
public static RecoveryPlan createRecoveryPlan(final ArrayList snapshots)
{
long lastLeadershipTermId = NULL_VALUE;
long lastTermBaseLogPosition = 0;
long committedLogPosition = 0;
long appendedLogPosition = 0;
final int snapshotStepsSize = snapshots.size();
if (snapshotStepsSize > 0)
{
final Snapshot snapshot = snapshots.get(0);
lastLeadershipTermId = snapshot.leadershipTermId;
lastTermBaseLogPosition = snapshot.termBaseLogPosition;
appendedLogPosition = snapshot.logPosition;
committedLogPosition = snapshot.logPosition;
}
return new RecoveryPlan(
lastLeadershipTermId,
lastTermBaseLogPosition,
appendedLogPosition,
committedLogPosition,
snapshots,
null);
}
/**
* Is the given leadershipTermId unknown for the log?
*
* @param leadershipTermId to check.
* @return true if term has not yet been appended otherwise false.
*/
public boolean isUnknown(final long leadershipTermId)
{
return NULL_VALUE == cacheIndexByLeadershipTermIdMap.get(leadershipTermId);
}
/**
* Append a log entry for a leadership term. Terms must be appended in ascending order.
*
* @param recordingId of the log in the archive.
* @param leadershipTermId for the appended term.
* @param termBaseLogPosition for the beginning of the term.
* @param timestamp at the beginning of the appended term.
*/
public void appendTerm(
final long recordingId, final long leadershipTermId, final long termBaseLogPosition, final long timestamp)
{
validateTermRecordingId(recordingId);
long logPosition = NULL_POSITION;
if (!entriesCache.isEmpty())
{
if (cacheIndexByLeadershipTermIdMap.containsKey(leadershipTermId))
{
throw new ClusterException("duplicate TERM entry for leadershipTermId=" + leadershipTermId);
}
final long previousLeadershipTermId = leadershipTermId - 1;
if (cacheIndexByLeadershipTermIdMap.containsKey(previousLeadershipTermId))
{
commitLogPosition(previousLeadershipTermId, termBaseLogPosition);
}
final Entry nextTermEntry = findTermEntry(leadershipTermId + 1);
if (null != nextTermEntry)
{
logPosition = nextTermEntry.termBaseLogPosition;
}
}
final int index = append(
ENTRY_TYPE_TERM,
recordingId,
leadershipTermId,
termBaseLogPosition,
logPosition,
timestamp,
NULL_VALUE, RECORDING_LOG_FILE_NAME);
cacheIndexByLeadershipTermIdMap.put(leadershipTermId, index);
}
/**
* Append a log entry for a snapshot. Snapshots must be for the current term.
*
* @param recordingId in the archive for the snapshot.
* @param leadershipTermId for the current term
* @param termBaseLogPosition at the beginning of the leadership term.
* @param logPosition within the current term or accumulated length for the log.
* @param timestamp at which the snapshot was taken.
* @param serviceId for which the snapshot is recorded.
*/
public void appendSnapshot(
final long recordingId,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long timestamp,
final int serviceId)
{
validateRecordingId(recordingId);
if (!restoreInvalidSnapshot(
recordingId, leadershipTermId, termBaseLogPosition, logPosition, timestamp, serviceId))
{
append(
ENTRY_TYPE_SNAPSHOT,
recordingId,
leadershipTermId,
termBaseLogPosition,
logPosition,
timestamp,
serviceId,
null);
}
}
/**
* Append a log entry for a snapshot. Snapshots must be for the current term.
*
* @param recordingId in the archive for the snapshot.
* @param leadershipTermId for the current term
* @param termBaseLogPosition at the beginning of the leadership term.
* @param logPosition within the current term or accumulated length for the log.
* @param timestamp at which the snapshot was taken.
* @param serviceId for which the snapshot is recorded.
* @param archiveEndpoint endpoint for the archive where
*/
public void appendStandbySnapshot(
final long recordingId,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long timestamp,
final int serviceId,
final String archiveEndpoint)
{
validateRecordingId(recordingId);
if (Strings.isEmpty(archiveEndpoint))
{
throw new ClusterException("Remote snapshots must has a valid endpoint");
}
if (!restoreInvalidSnapshot(
recordingId, leadershipTermId, termBaseLogPosition, logPosition, timestamp, serviceId))
{
append(
ENTRY_TYPE_STANDBY_SNAPSHOT,
recordingId,
leadershipTermId,
termBaseLogPosition,
logPosition,
timestamp,
serviceId,
archiveEndpoint);
}
}
/**
* Commit the log position reached in a leadership term.
*
* @param leadershipTermId for committing the term position reached.
* @param logPosition reached in the leadership term.
*/
public void commitLogPosition(final long leadershipTermId, final long logPosition)
{
final int index = (int)cacheIndexByLeadershipTermIdMap.get(leadershipTermId);
if (NULL_VALUE == index)
{
throw new ClusterException("unknown leadershipTermId=" + leadershipTermId);
}
final Entry entry = entriesCache.get(index);
if (entry.logPosition != logPosition)
{
buffer.putLong(0, logPosition, LITTLE_ENDIAN);
persistToStorage(entry.position, LOG_POSITION_OFFSET, SIZE_OF_LONG);
entriesCache.set(index, entry.logPosition(logPosition));
}
}
/**
* Invalidate an entry in the log, so it is no longer valid. Be careful that the recording log is not left in an
* invalid state for recovery.
*
* @param leadershipTermId to match for validation.
* @param entryIndex reached in the leadership term.
* @see #invalidateLatestSnapshot()
*/
public void invalidateEntry(final long leadershipTermId, final int entryIndex)
{
Entry invalidEntry = null;
for (int i = entriesCache.size() - 1; i >= 0; i--)
{
final Entry entry = entriesCache.get(i);
if (entry.leadershipTermId == leadershipTermId && entry.entryIndex == entryIndex)
{
invalidEntry = entry.invalidate();
entriesCache.set(i, invalidEntry);
if (ENTRY_TYPE_TERM == entry.type)
{
cacheIndexByLeadershipTermIdMap.remove(leadershipTermId);
}
else if (ENTRY_TYPE_SNAPSHOT == entry.type)
{
invalidSnapshots.add(i);
}
break;
}
}
if (null == invalidEntry)
{
throw new ClusterException("unknown entry index: " + entryIndex);
}
final int invalidEntryType = ENTRY_TYPE_INVALID_FLAG | invalidEntry.type;
buffer.putInt(0, invalidEntryType, LITTLE_ENDIAN);
persistToStorage(invalidEntry.position, ENTRY_TYPE_OFFSET, SIZE_OF_INT);
}
/**
* Remove an entry in the log and reload the caches.
*
* @param leadershipTermId to match for validation.
* @param entryIndex reached in the leadership term.
*/
void removeEntry(final long leadershipTermId, final int entryIndex)
{
Entry entryToRemove = null;
for (int i = entriesCache.size() - 1; i >= 0; i--)
{
final Entry entry = entriesCache.get(i);
if (entry.leadershipTermId == leadershipTermId && entry.entryIndex == entryIndex)
{
entryToRemove = entry;
break;
}
}
if (null == entryToRemove)
{
throw new ClusterException("unknown entry index: " + entryIndex);
}
buffer.putInt(0, NULL_VALUE, LITTLE_ENDIAN);
persistToStorage(entryToRemove.position, ENTRY_TYPE_OFFSET, SIZE_OF_INT);
reload();
}
/**
* Return a collection of the most recent remote snapshots grouped by the archiveEndpoint where the snapshot is
* stored.
*
* @param serviceCount to ensure that we have a complete set of snapshots.
* @return collection of snapshots.
*/
public Map> latestStandbySnapshots(final int serviceCount)
{
final Map> latestStandbySnapshots = new Object2ObjectHashMap<>();
final Map>> standbySnapshots = new Object2ObjectHashMap<>();
for (int i = entriesCache.size() - 1; i >= 0; i--)
{
final Entry entry = entriesCache.get(i);
if (ENTRY_TYPE_STANDBY_SNAPSHOT == entry.type)
{
standbySnapshots.computeIfAbsent(entry.archiveEndpoint, (s) -> new TreeMap<>())
.computeIfAbsent(entry.logPosition, (l) -> new ArrayList<>())
.add(entry);
}
}
standbySnapshots.forEach(
(k, v) ->
{
while (!v.isEmpty())
{
final Map.Entry> lastEntry = v.lastEntry();
final int snapshotCount = serviceCount + 1;
if (lastEntry.getValue().size() == snapshotCount)
{
latestStandbySnapshots.put(k, lastEntry.getValue());
break;
}
else
{
v.remove(lastEntry.getKey());
}
}
});
return latestStandbySnapshots;
}
/**
* {@inheritDoc}
*/
public String toString()
{
return "RecordingLog{" +
"entries=" + entriesCache +
", cacheIndex=" + cacheIndexByLeadershipTermIdMap +
'}';
}
static String typeAsString(final int type)
{
switch (type)
{
case ENTRY_TYPE_TERM:
return "TERM";
case ENTRY_TYPE_SNAPSHOT:
return "SNAPSHOT";
case ENTRY_TYPE_STANDBY_SNAPSHOT:
return "STANDBY_SNAPSHOT";
default:
return "UNKNOWN";
}
}
static void addSnapshots(
final ArrayList snapshots,
final ArrayList entries,
final int serviceCount,
final int snapshotIndex)
{
final Entry snapshot = entries.get(snapshotIndex);
snapshots.add(new Snapshot(
snapshot.recordingId,
snapshot.leadershipTermId,
snapshot.termBaseLogPosition,
snapshot.logPosition,
snapshot.timestamp,
snapshot.serviceId));
for (int i = 1; i <= serviceCount; i++)
{
if ((snapshotIndex - i) < 0)
{
throw new ClusterException("snapshot missing for service at index " + i + " in " + entries);
}
final Entry entry = entries.get(snapshotIndex - i);
if (ENTRY_TYPE_SNAPSHOT == entry.type &&
entry.leadershipTermId == snapshot.leadershipTermId &&
entry.logPosition == snapshot.logPosition)
{
snapshots.add(entry.serviceId + 1, new Snapshot(
entry.recordingId,
entry.leadershipTermId,
entry.termBaseLogPosition,
entry.logPosition,
entry.timestamp,
entry.serviceId));
}
}
}
static void writeEntryToBuffer(final Entry entry, final UnsafeBuffer buffer)
{
buffer.putLong(RECORDING_ID_OFFSET, entry.recordingId, LITTLE_ENDIAN);
buffer.putLong(LEADERSHIP_TERM_ID_OFFSET, entry.leadershipTermId, LITTLE_ENDIAN);
buffer.putLong(TERM_BASE_LOG_POSITION_OFFSET, entry.termBaseLogPosition, LITTLE_ENDIAN);
buffer.putLong(LOG_POSITION_OFFSET, entry.logPosition, LITTLE_ENDIAN);
buffer.putLong(TIMESTAMP_OFFSET, entry.timestamp, LITTLE_ENDIAN);
buffer.putInt(SERVICE_ID_OFFSET, entry.serviceId, LITTLE_ENDIAN);
buffer.putInt(ENTRY_TYPE_OFFSET, entry.type, LITTLE_ENDIAN);
if (!Strings.isEmpty(entry.archiveEndpoint))
{
buffer.putStringAscii(ENDPOINT_OFFSET, entry.archiveEndpoint);
}
}
static boolean isValidSnapshot(final Entry entry)
{
return entry.isValid && ENTRY_TYPE_SNAPSHOT == entry.type;
}
static boolean isInvalidSnapshot(final Entry entry)
{
return !entry.isValid && ENTRY_TYPE_SNAPSHOT == entry.type;
}
void ensureCoherent(
final long recordingId,
final long initialLogLeadershipTermId,
final long initialTermBaseLogPosition,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long nowNs,
final long timestamp,
final int fileSyncLevel)
{
Entry lastTerm = findLastTerm();
if (null == lastTerm)
{
for (long termId = max(0, initialLogLeadershipTermId); termId < leadershipTermId; termId++)
{
appendTerm(recordingId, termId, initialTermBaseLogPosition, timestamp);
}
appendTerm(recordingId, leadershipTermId, termBaseLogPosition, timestamp);
if (NULL_VALUE != logPosition)
{
commitLogPosition(leadershipTermId, logPosition);
}
}
else if (lastTerm.leadershipTermId < leadershipTermId)
{
if (NULL_VALUE == lastTerm.logPosition)
{
if (NULL_VALUE == termBaseLogPosition)
{
throw new ClusterException(
"Prior term was not committed: " + lastTerm +
" and logTermBasePosition was not specified: leadershipTermId = " + leadershipTermId +
", logTermBasePosition = " + termBaseLogPosition + ", logPosition = " + logPosition +
", nowNs = " + nowNs);
}
else
{
commitLogPosition(lastTerm.leadershipTermId, termBaseLogPosition);
lastTerm = Objects.requireNonNull(findLastTerm());
}
}
for (long termId = lastTerm.leadershipTermId + 1; termId < leadershipTermId; termId++)
{
appendTerm(recordingId, termId, lastTerm.logPosition, timestamp);
}
appendTerm(recordingId, leadershipTermId, lastTerm.logPosition, timestamp);
if (NULL_VALUE != logPosition)
{
commitLogPosition(leadershipTermId, logPosition);
}
}
else
{
if (NULL_VALUE != logPosition)
{
commitLogPosition(leadershipTermId, logPosition);
}
}
force(fileSyncLevel);
}
private static void validateRecordingId(final long recordingId)
{
if (NULL_VALUE == recordingId)
{
throw new ClusterException("invalid recordingId=" + recordingId);
}
}
private void validateTermRecordingId(final long recordingId)
{
validateRecordingId(recordingId);
if (recordingId != termRecordingId)
{
if (NULL_VALUE == termRecordingId)
{
termRecordingId = recordingId;
}
else
{
throw new ClusterException(
"invalid TERM recordingId=" + recordingId + ", expected recordingId=" + termRecordingId);
}
}
}
private boolean restoreInvalidSnapshot(
final long recordingId,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long timestamp,
final int serviceId)
{
for (int i = invalidSnapshots.size() - 1; i >= 0; i--)
{
final int entryCacheIndex = invalidSnapshots.getInt(i);
final Entry entry = entriesCache.get(entryCacheIndex);
if (matchesEntry(entry, leadershipTermId, termBaseLogPosition, logPosition, serviceId))
{
final Entry validatedEntry = new Entry(
recordingId,
leadershipTermId,
termBaseLogPosition,
logPosition,
timestamp,
serviceId,
ENTRY_TYPE_SNAPSHOT,
null,
true,
entry.position,
entry.entryIndex);
writeEntryToBuffer(validatedEntry, buffer);
persistToStorage(entry, buffer);
entriesCache.set(entryCacheIndex, validatedEntry);
invalidSnapshots.fastUnorderedRemove(i);
return true;
}
}
return false;
}
private int append(
final int entryType,
final long recordingId,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final long timestamp,
final int serviceId,
final String endpoint)
{
final Entry entry = new Entry(
recordingId,
leadershipTermId,
termBaseLogPosition,
logPosition,
timestamp,
serviceId,
entryType,
endpoint,
true,
NULL_VALUE,
nextEntryIndex);
writeEntryToBuffer(entry, buffer);
persistToStorage(entry, buffer);
nextEntryIndex++;
final ArrayList entries = this.entriesCache;
final int size = entries.size();
int index = size;
for (int i = size - 1; i >= 0; i--)
{
final Entry e = entries.get(i);
if (ENTRY_COMPARATOR.compare(entry, e) < 0)
{
index--;
}
else
{
break;
}
}
if (index < size)
{
entries.add(null);
for (int i = size - 1; i >= index; i--)
{
entries.set(i + 1, entries.get(i));
}
entries.set(index, entry);
final Long2LongHashMap.EntryIterator entryIterator = cacheIndexByLeadershipTermIdMap.entrySet().iterator();
while (entryIterator.hasNext())
{
entryIterator.next();
if (entryIterator.getLongKey() > leadershipTermId)
{
entryIterator.setValue(entryIterator.getLongValue() + 1);
}
}
for (int i = invalidSnapshots.size() - 1; i >= 0; i--)
{
final int snapshotIndex = invalidSnapshots.getInt(i);
if (snapshotIndex >= index)
{
invalidSnapshots.set(i, snapshotIndex + 1);
}
else
{
break;
}
}
}
else
{
entries.add(entry);
}
return index;
}
private void persistToStorage(final long entryPosition, final int offset, final int length)
{
byteBuffer.limit(length).position(0);
final long position = entryPosition + offset;
try
{
if (length != fileChannel.write(byteBuffer, position))
{
throw new ClusterException("failed to write field atomically");
}
}
catch (final Exception ex)
{
LangUtil.rethrowUnchecked(ex);
}
}
private void persistToStorage(final Entry entry, final DirectBuffer directBuffer)
{
final ByteBuffer byteBuffer = directBuffer.byteBuffer();
byteBuffer.limit(entry.length()).position(0);
try
{
if (NULL_VALUE == entry.position)
{
final long nextPosition = align(fileChannel.size(), CACHE_LINE_LENGTH);
entry.position(nextPosition);
}
if (entry.length() != fileChannel.write(byteBuffer, entry.position))
{
throw new ClusterException("failed to write field atomically");
}
}
catch (final Exception ex)
{
LangUtil.rethrowUnchecked(ex);
}
}
private int captureEntriesFromBuffer(
final long filePosition, final ByteBuffer byteBuffer, final UnsafeBuffer buffer, final ArrayList entries)
{
int consumed = 0; // This is assuming that we have filled the byte buffer from the beginning.
for (int length = byteBuffer.limit(); consumed < length;)
{
final long position = filePosition + consumed;
final int entryType = buffer.getInt(consumed + ENTRY_TYPE_OFFSET);
final int type = entryType & ~ENTRY_TYPE_INVALID_FLAG;
final boolean isValid = (entryType & ENTRY_TYPE_INVALID_FLAG) == 0;
final Entry entry = new Entry(
buffer.getLong(consumed + RECORDING_ID_OFFSET, LITTLE_ENDIAN),
buffer.getLong(consumed + LEADERSHIP_TERM_ID_OFFSET, LITTLE_ENDIAN),
buffer.getLong(consumed + TERM_BASE_LOG_POSITION_OFFSET, LITTLE_ENDIAN),
buffer.getLong(consumed + LOG_POSITION_OFFSET, LITTLE_ENDIAN),
buffer.getLong(consumed + TIMESTAMP_OFFSET, LITTLE_ENDIAN),
buffer.getInt(consumed + SERVICE_ID_OFFSET, LITTLE_ENDIAN),
type,
(ENTRY_TYPE_STANDBY_SNAPSHOT == type) ? buffer.getStringAscii(consumed + ENDPOINT_OFFSET) : null,
isValid,
position,
nextEntryIndex);
if (NULL_VALUE != entryType)
{
if (ENTRY_TYPE_TERM == type)
{
validateTermRecordingId(entry.recordingId);
}
entries.add(entry);
}
consumed += entry.length();
++nextEntryIndex;
}
byteBuffer.position(consumed);
return consumed;
}
private static void syncDirectory(final File dir)
{
try (FileChannel fileChannel = FileChannel.open(dir.toPath()))
{
fileChannel.force(true);
}
catch (final IOException ignore)
{
}
}
private void commitEntryLogPosition(final int entryIndex, final long value)
{
buffer.putLong(0, value, LITTLE_ENDIAN);
persistToStorage(entryIndex, LOG_POSITION_OFFSET, SIZE_OF_LONG);
}
private static void planRecovery(
final ArrayList snapshots,
final MutableReference logRef,
final ArrayList entries,
final AeronArchive archive,
final int serviceCount,
final long replicatedRecordingId)
{
if (entries.isEmpty())
{
if (Aeron.NULL_VALUE != replicatedRecordingId)
{
final RecordingExtent recordingExtent = new RecordingExtent();
if (archive.listRecording(replicatedRecordingId, recordingExtent) == 0)
{
throw new ClusterException("unknown recording id: " + replicatedRecordingId);
}
logRef.set(new Log(
replicatedRecordingId,
NULL_VALUE,
0,
0,
recordingExtent.startPosition,
recordingExtent.stopPosition,
recordingExtent.initialTermId,
recordingExtent.termBufferLength,
recordingExtent.mtuLength,
recordingExtent.sessionId));
}
return;
}
int logIndex = -1;
int snapshotIndex = -1;
for (int i = entries.size() - 1; i >= 0; i--)
{
final Entry entry = entries.get(i);
if (-1 == snapshotIndex && isValidSnapshot(entry) &&
entry.serviceId == ConsensusModule.Configuration.SERVICE_ID)
{
snapshotIndex = i;
}
else if (-1 == logIndex &&
isValidTerm(entry) &&
NULL_VALUE != entry.recordingId)
{
logIndex = i;
}
else if (-1 != snapshotIndex && -1 != logIndex)
{
break;
}
}
if (-1 != snapshotIndex)
{
addSnapshots(snapshots, entries, serviceCount, snapshotIndex);
}
if (-1 != logIndex)
{
final Entry entry = entries.get(logIndex);
final RecordingExtent recordingExtent = new RecordingExtent();
if (archive.listRecording(entry.recordingId, recordingExtent) == 0)
{
throw new ClusterException("unknown recording id: " + entry.recordingId);
}
final long startPosition = -1 == snapshotIndex ?
recordingExtent.startPosition : snapshots.get(0).logPosition;
logRef.set(new Log(
entry.recordingId,
entry.leadershipTermId,
entry.termBaseLogPosition,
entry.logPosition,
startPosition,
recordingExtent.stopPosition,
recordingExtent.initialTermId,
recordingExtent.termBufferLength,
recordingExtent.mtuLength,
recordingExtent.sessionId));
}
}
private static boolean isValidTerm(final Entry entry)
{
return ENTRY_TYPE_TERM == entry.type && entry.isValid;
}
private static boolean matchesEntry(
final Entry entry,
final long leadershipTermId,
final long termBaseLogPosition,
final long logPosition,
final int serviceId)
{
return entry.leadershipTermId == leadershipTermId &&
entry.termBaseLogPosition == termBaseLogPosition &&
entry.logPosition == logPosition &&
entry.serviceId == serviceId;
}
}