private static void logMutationSize(HTableInterface htable, List<Mutation> mutations) {
long byteSize = 0;
int keyValueCount = 0;
for (Mutation mutation : mutations) {
if (mutation.getFamilyCellMap() != null) { // Not a Delete of the row
for (Entry<byte[], List<Cell>> entry : mutation.getFamilyCellMap().entrySet()) {
if (entry.getValue() != null) {
for (Cell kv : entry.getValue()) {
byteSize += CellUtil.estimatedSizeOf(kv);
keyValueCount++;
}
}
}
}
}
logger.debug(
"Sending "
+ mutations.size()
+ " mutations for "
+ Bytes.toString(htable.getTableName())
+ " with "
+ keyValueCount
+ " key values of total size "
+ byteSize
+ " bytes");
}
/**
* Helper method for a {@link KeyValueBuilder} that catches an IOException from a {@link Put} when
* adding a {@link KeyValue} generated by the KeyValueBuilder.
*
* @throws RuntimeException if there is an IOException thrown from the underlying {@link Put}
*/
@SuppressWarnings("javadoc")
public static void addQuietly(Mutation m, KeyValueBuilder builder, KeyValue kv) {
byte[] family = CellUtil.cloneFamily(kv);
List<Cell> list = m.getFamilyCellMap().get(family);
if (list == null) {
list = new ArrayList<Cell>();
m.getFamilyCellMap().put(family, list);
}
list.add(kv);
}
public static ScanRanges newScanRanges(List<Mutation> mutations) throws SQLException {
List<KeyRange> keys = Lists.newArrayListWithExpectedSize(mutations.size());
for (Mutation m : mutations) {
keys.add(PVarbinary.INSTANCE.getKeyRange(m.getRow()));
}
ScanRanges keyRanges =
ScanRanges.create(
SchemaUtil.VAR_BINARY_SCHEMA,
Collections.singletonList(keys),
ScanUtil.SINGLE_COLUMN_SLOT_SPAN);
return keyRanges;
}
private static byte[] getTableName(List<Mutation> tableMetaData, byte[] physicalTableName) {
if (physicalTableName != null) {
return physicalTableName;
}
byte[][] rowKeyMetadata = new byte[3][];
Mutation m = MetaDataUtil.getTableHeaderRow(tableMetaData);
byte[] key = m.getRow();
SchemaUtil.getVarChars(key, rowKeyMetadata);
byte[] schemaBytes = rowKeyMetadata[PhoenixDatabaseMetaData.SCHEMA_NAME_INDEX];
byte[] tableBytes = rowKeyMetadata[PhoenixDatabaseMetaData.TABLE_NAME_INDEX];
return SchemaUtil.getTableNameAsBytes(schemaBytes, tableBytes);
}
private static void commitBatch(HRegion region, List<Mutation> mutations, byte[] indexUUID)
throws IOException {
if (indexUUID != null) {
for (Mutation m : mutations) {
m.setAttribute(PhoenixIndexCodec.INDEX_UUID, indexUUID);
}
}
@SuppressWarnings("unchecked")
Mutation[] mutationArray = new Mutation[mutations.size()];
// TODO: should we use the one that is all or none?
region.batchMutate(mutations.toArray(mutationArray));
}
private void mutateRow(Mutation mut, Integer lockId) throws IOException {
@SuppressWarnings("unchecked")
Pair<Mutation, Integer> pair[] = new Pair[1];
mut.setWriteToWAL(true);
pair[0] = new Pair<Mutation, Integer>(mut, lockId);
region.batchMutate(pair);
}
private void removeIfPresent(Mutation m, byte[] family, byte[] qualifier) {
Map<byte[], List<KeyValue>> familyMap = m.getFamilyMap();
List<KeyValue> kvs = familyMap.get(family);
if (kvs != null) {
Iterator<KeyValue> iterator = kvs.iterator();
while (iterator.hasNext()) {
KeyValue kv = iterator.next();
if (Bytes.compareTo(kv.getQualifier(), qualifier) == 0) {
iterator.remove();
}
}
}
}
@Override
public Result getCurrentRowState(Mutation m, Collection<? extends ColumnReference> columns)
throws IOException {
byte[] row = m.getRow();
// need to use a scan here so we can get raw state, which Get doesn't provide.
Scan s = IndexManagementUtil.newLocalStateScan(Collections.singletonList(columns));
s.setStartRow(row);
s.setStopRow(row);
HRegion region = this.env.getRegion();
RegionScanner scanner = region.getScanner(s);
List<KeyValue> kvs = new ArrayList<KeyValue>(1);
boolean more = scanner.next(kvs);
assert !more : "Got more than one result when scanning" + " a single row in the primary table!";
Result r = new Result(kvs);
scanner.close();
return r;
}
@SuppressWarnings("deprecation")
public void commit() throws SQLException {
int i = 0;
byte[] tenantId = connection.getTenantId() == null ? null : connection.getTenantId().getBytes();
long[] serverTimeStamps = validate();
Iterator<Map.Entry<TableRef, Map<ImmutableBytesPtr, Map<PColumn, byte[]>>>> iterator =
this.mutations.entrySet().iterator();
List<Map.Entry<TableRef, Map<ImmutableBytesPtr, Map<PColumn, byte[]>>>> committedList =
Lists.newArrayListWithCapacity(this.mutations.size());
// add tracing for this operation
TraceScope trace = Tracing.startNewSpan(connection, "Committing mutations to tables");
Span span = trace.getSpan();
while (iterator.hasNext()) {
Map.Entry<TableRef, Map<ImmutableBytesPtr, Map<PColumn, byte[]>>> entry = iterator.next();
Map<ImmutableBytesPtr, Map<PColumn, byte[]>> valuesMap = entry.getValue();
TableRef tableRef = entry.getKey();
PTable table = tableRef.getTable();
table.getIndexMaintainers(tempPtr);
boolean hasIndexMaintainers = tempPtr.getLength() > 0;
boolean isDataTable = true;
long serverTimestamp = serverTimeStamps[i++];
Iterator<Pair<byte[], List<Mutation>>> mutationsIterator =
addRowMutations(tableRef, valuesMap, serverTimestamp, false);
while (mutationsIterator.hasNext()) {
Pair<byte[], List<Mutation>> pair = mutationsIterator.next();
byte[] htableName = pair.getFirst();
List<Mutation> mutations = pair.getSecond();
// create a span per target table
// TODO maybe we can be smarter about the table name to string here?
Span child =
Tracing.child(span, "Writing mutation batch for table: " + Bytes.toString(htableName));
int retryCount = 0;
boolean shouldRetry = false;
do {
ServerCache cache = null;
if (hasIndexMaintainers && isDataTable) {
byte[] attribValue = null;
byte[] uuidValue;
if (IndexMetaDataCacheClient.useIndexMetadataCache(
connection, mutations, tempPtr.getLength())) {
IndexMetaDataCacheClient client = new IndexMetaDataCacheClient(connection, tableRef);
cache = client.addIndexMetadataCache(mutations, tempPtr);
child.addTimelineAnnotation("Updated index metadata cache");
uuidValue = cache.getId();
// If we haven't retried yet, retry for this case only, as it's possible that
// a split will occur after we send the index metadata cache to all known
// region servers.
shouldRetry = true;
} else {
attribValue = ByteUtil.copyKeyBytesIfNecessary(tempPtr);
uuidValue = ServerCacheClient.generateId();
}
// Either set the UUID to be able to access the index metadata from the cache
// or set the index metadata directly on the Mutation
for (Mutation mutation : mutations) {
if (tenantId != null) {
mutation.setAttribute(PhoenixRuntime.TENANT_ID_ATTRIB, tenantId);
}
mutation.setAttribute(PhoenixIndexCodec.INDEX_UUID, uuidValue);
if (attribValue != null) {
mutation.setAttribute(PhoenixIndexCodec.INDEX_MD, attribValue);
}
}
}
SQLException sqlE = null;
HTableInterface hTable = connection.getQueryServices().getTable(htableName);
try {
if (logger.isDebugEnabled()) logMutationSize(hTable, mutations);
long startTime = System.currentTimeMillis();
child.addTimelineAnnotation("Attempt " + retryCount);
hTable.batch(mutations);
child.stop();
shouldRetry = false;
if (logger.isDebugEnabled())
logger.debug(
"Total time for batch call of "
+ mutations.size()
+ " mutations into "
+ table.getName().getString()
+ ": "
+ (System.currentTimeMillis() - startTime)
+ " ms");
committedList.add(entry);
} catch (Exception e) {
SQLException inferredE = ServerUtil.parseServerExceptionOrNull(e);
if (inferredE != null) {
if (shouldRetry
&& retryCount == 0
&& inferredE.getErrorCode()
== SQLExceptionCode.INDEX_METADATA_NOT_FOUND.getErrorCode()) {
// Swallow this exception once, as it's possible that we split after sending the
// index metadata
// and one of the region servers doesn't have it. This will cause it to have it the
// next go around.
// If it fails again, we don't retry.
String msg =
"Swallowing exception and retrying after clearing meta cache on connection. "
+ inferredE;
logger.warn(msg);
connection.getQueryServices().clearTableRegionCache(htableName);
// add a new child span as this one failed
child.addTimelineAnnotation(msg);
child.stop();
child = Tracing.child(span, "Failed batch, attempting retry");
continue;
}
e = inferredE;
}
// Throw to client with both what was committed so far and what is left to be committed.
// That way, client can either undo what was done or try again with what was not done.
sqlE =
new CommitException(
e,
this,
new MutationState(
committedList, this.sizeOffset, this.maxSize, this.connection));
} finally {
try {
hTable.close();
} catch (IOException e) {
if (sqlE != null) {
sqlE.setNextException(ServerUtil.parseServerException(e));
} else {
sqlE = ServerUtil.parseServerException(e);
}
} finally {
try {
if (cache != null) {
cache.close();
}
} finally {
if (sqlE != null) {
throw sqlE;
}
}
}
}
} while (shouldRetry && retryCount++ < 1);
isDataTable = false;
}
numRows -= entry.getValue().size();
iterator.remove(); // Remove batches as we process them
}
trace.close();
assert (numRows == 0);
assert (this.mutations.isEmpty());
}
/**
* Helper to add a {@link Mutation} to the values stored for the current row
*
* @param pendingUpdate update to apply
*/
public void addUpdateForTesting(Mutation pendingUpdate) {
for (Map.Entry<byte[], List<KeyValue>> e : pendingUpdate.getFamilyMap().entrySet()) {
List<KeyValue> edits = e.getValue();
addUpdate(edits);
}
}
/**
* Prepare the regions and region files.
*
* @param server Hosting server instance. Can be null when testing (won't try and update in zk if
* a null server)
* @param services Used to online/offline regions.
* @param user
* @throws IOException If thrown, transaction failed. Call {@link #rollback(Server,
* RegionServerServices)}
* @return Regions created
*/
@VisibleForTesting
PairOfSameType<Region> createDaughters(
final Server server, final RegionServerServices services, User user) throws IOException {
LOG.info("Starting split of region " + this.parent);
if ((server != null && server.isStopped()) || (services != null && services.isStopping())) {
throw new IOException("Server is stopped or stopping");
}
assert !this.parent.lock.writeLock().isHeldByCurrentThread()
: "Unsafe to hold write lock while performing RPCs";
transition(SplitTransactionPhase.BEFORE_PRE_SPLIT_HOOK);
// Coprocessor callback
if (this.parent.getCoprocessorHost() != null) {
// TODO: Remove one of these
parent.getCoprocessorHost().preSplit(user);
parent.getCoprocessorHost().preSplit(splitrow, user);
}
transition(SplitTransactionPhase.AFTER_PRE_SPLIT_HOOK);
// If true, no cluster to write meta edits to or to update znodes in.
boolean testing =
server == null
? true
: server.getConfiguration().getBoolean("hbase.testing.nocluster", false);
this.fileSplitTimeout =
testing
? this.fileSplitTimeout
: server
.getConfiguration()
.getLong("hbase.regionserver.fileSplitTimeout", this.fileSplitTimeout);
PairOfSameType<Region> daughterRegions = stepsBeforePONR(server, services, testing);
final List<Mutation> metaEntries = new ArrayList<Mutation>();
boolean ret = false;
if (this.parent.getCoprocessorHost() != null) {
ret = parent.getCoprocessorHost().preSplitBeforePONR(splitrow, metaEntries, user);
if (ret) {
throw new IOException(
"Coprocessor bypassing region "
+ parent.getRegionInfo().getRegionNameAsString()
+ " split.");
}
try {
for (Mutation p : metaEntries) {
HRegionInfo.parseRegionName(p.getRow());
}
} catch (IOException e) {
LOG.error(
"Row key of mutation from coprossor is not parsable as region name."
+ "Mutations from coprocessor should only for hbase:meta table.");
throw e;
}
}
// This is the point of no return. Adding subsequent edits to .META. as we
// do below when we do the daughter opens adding each to .META. can fail in
// various interesting ways the most interesting of which is a timeout
// BUT the edits all go through (See HBASE-3872). IF we reach the PONR
// then subsequent failures need to crash out this regionserver; the
// server shutdown processing should be able to fix-up the incomplete split.
// The offlined parent will have the daughters as extra columns. If
// we leave the daughter regions in place and do not remove them when we
// crash out, then they will have their references to the parent in place
// still and the server shutdown fixup of .META. will point to these
// regions.
// We should add PONR JournalEntry before offlineParentInMeta,so even if
// OfflineParentInMeta timeout,this will cause regionserver exit,and then
// master ServerShutdownHandler will fix daughter & avoid data loss. (See
// HBase-4562).
transition(SplitTransactionPhase.PONR);
// Edit parent in meta. Offlines parent region and adds splita and splitb
// as an atomic update. See HBASE-7721. This update to META makes the region
// will determine whether the region is split or not in case of failures.
// If it is successful, master will roll-forward, if not, master will rollback
// and assign the parent region.
if (services != null
&& !services.reportRegionStateTransition(
TransitionCode.SPLIT_PONR, parent.getRegionInfo(), hri_a, hri_b)) {
// Passed PONR, let SSH clean it up
throw new IOException(
"Failed to notify master that split passed PONR: "
+ parent.getRegionInfo().getRegionNameAsString());
}
return daughterRegions;
}
/**
* Prepare the merged region and region files.
*
* @param server Hosting server instance. Can be null when testing
* @param services Used to online/offline regions.
* @return merged region
* @throws IOException If thrown, transaction failed. Call {@link #rollback(Server,
* RegionServerServices)}
*/
HRegion createMergedRegion(final Server server, final RegionServerServices services)
throws IOException {
LOG.info(
"Starting merge of "
+ region_a
+ " and "
+ region_b.getRegionNameAsString()
+ ", forcible="
+ forcible);
if ((server != null && server.isStopped()) || (services != null && services.isStopping())) {
throw new IOException("Server is stopped or stopping");
}
if (rsCoprocessorHost != null) {
if (rsCoprocessorHost.preMerge(this.region_a, this.region_b)) {
throw new IOException(
"Coprocessor bypassing regions " + this.region_a + " " + this.region_b + " merge.");
}
}
// If true, no cluster to write meta edits to or to use coordination.
boolean testing =
server == null
? true
: server.getConfiguration().getBoolean("hbase.testing.nocluster", false);
HRegion mergedRegion = stepsBeforePONR(server, services, testing);
@MetaMutationAnnotation List<Mutation> metaEntries = new ArrayList<Mutation>();
if (rsCoprocessorHost != null) {
if (rsCoprocessorHost.preMergeCommit(this.region_a, this.region_b, metaEntries)) {
throw new IOException(
"Coprocessor bypassing regions " + this.region_a + " " + this.region_b + " merge.");
}
try {
for (Mutation p : metaEntries) {
HRegionInfo.parseRegionName(p.getRow());
}
} catch (IOException e) {
LOG.error(
"Row key of mutation from coprocessor is not parsable as region name."
+ "Mutations from coprocessor should only be for hbase:meta table.",
e);
throw e;
}
}
// This is the point of no return. Similar with SplitTransaction.
// IF we reach the PONR then subsequent failures need to crash out this
// regionserver
this.journal.add(JournalEntry.PONR);
// Add merged region and delete region_a and region_b
// as an atomic update. See HBASE-7721. This update to hbase:meta makes the region
// will determine whether the region is merged or not in case of failures.
// If it is successful, master will roll-forward, if not, master will
// rollback
if (!testing && useCoordinationForAssignment) {
if (metaEntries.isEmpty()) {
MetaTableAccessor.mergeRegions(
server.getConnection(),
mergedRegion.getRegionInfo(),
region_a.getRegionInfo(),
region_b.getRegionInfo(),
server.getServerName(),
region_a.getTableDesc().getRegionReplication());
} else {
mergeRegionsAndPutMetaEntries(
server.getConnection(),
mergedRegion.getRegionInfo(),
region_a.getRegionInfo(),
region_b.getRegionInfo(),
server.getServerName(),
metaEntries,
region_a.getTableDesc().getRegionReplication());
}
} else if (services != null && !useCoordinationForAssignment) {
if (!services.reportRegionStateTransition(
TransitionCode.MERGE_PONR,
mergedRegionInfo,
region_a.getRegionInfo(),
region_b.getRegionInfo())) {
// Passed PONR, let SSH clean it up
throw new IOException(
"Failed to notify master that merge passed PONR: "
+ region_a.getRegionInfo().getRegionNameAsString()
+ " and "
+ region_b.getRegionInfo().getRegionNameAsString());
}
}
return mergedRegion;
}
/**
* Override the preAppend for checkAndPut and checkAndDelete, as we need the ability to a) set the
* TimeRange for the Get being done and b) return something back to the client to indicate
* success/failure
*/
@SuppressWarnings("deprecation")
@Override
public Result preAppend(
final ObserverContext<RegionCoprocessorEnvironment> e, final Append append)
throws IOException {
byte[] opBuf = append.getAttribute(OPERATION_ATTRIB);
if (opBuf == null) {
return null;
}
Op op = Op.values()[opBuf[0]];
long clientTimestamp = HConstants.LATEST_TIMESTAMP;
byte[] clientTimestampBuf = append.getAttribute(MAX_TIMERANGE_ATTRIB);
if (clientTimestampBuf != null) {
clientTimestamp = Bytes.toLong(clientTimestampBuf);
}
boolean hadClientTimestamp = (clientTimestamp != HConstants.LATEST_TIMESTAMP);
if (hadClientTimestamp) {
// Prevent race condition of creating two sequences at the same timestamp
// by looking for a sequence at or after the timestamp at which it'll be
// created.
if (op == Op.CREATE_SEQUENCE) {
clientTimestamp++;
}
} else {
clientTimestamp = EnvironmentEdgeManager.currentTimeMillis();
clientTimestampBuf = Bytes.toBytes(clientTimestamp);
}
RegionCoprocessorEnvironment env = e.getEnvironment();
// We need to set this to prevent region.append from being called
e.bypass();
e.complete();
HRegion region = env.getRegion();
byte[] row = append.getRow();
region.startRegionOperation();
try {
Integer lid = region.getLock(null, row, true);
try {
KeyValue keyValue = append.getFamilyMap().values().iterator().next().iterator().next();
byte[] family = keyValue.getFamily();
byte[] qualifier = keyValue.getQualifier();
Get get = new Get(row);
get.setTimeRange(MetaDataProtocol.MIN_TABLE_TIMESTAMP, clientTimestamp);
get.addColumn(family, qualifier);
Result result = region.get(get);
if (result.isEmpty()) {
if (op == Op.DROP_SEQUENCE || op == Op.RESET_SEQUENCE) {
return getErrorResult(
row, clientTimestamp, SQLExceptionCode.SEQUENCE_UNDEFINED.getErrorCode());
}
} else {
if (op == Op.CREATE_SEQUENCE) {
return getErrorResult(
row, clientTimestamp, SQLExceptionCode.SEQUENCE_ALREADY_EXIST.getErrorCode());
}
}
Mutation m = null;
switch (op) {
case RESET_SEQUENCE:
KeyValue currentValueKV = result.raw()[0];
long expectedValue =
PDataType.LONG
.getCodec()
.decodeLong(append.getAttribute(CURRENT_VALUE_ATTRIB), 0, null);
long value =
PDataType.LONG
.getCodec()
.decodeLong(currentValueKV.getBuffer(), currentValueKV.getValueOffset(), null);
// Timestamp should match exactly, or we may have the wrong sequence
if (expectedValue != value || currentValueKV.getTimestamp() != clientTimestamp) {
return new Result(
Collections.singletonList(
KeyValueUtil.newKeyValue(
row,
PhoenixDatabaseMetaData.SEQUENCE_FAMILY_BYTES,
QueryConstants.EMPTY_COLUMN_BYTES,
currentValueKV.getTimestamp(),
ByteUtil.EMPTY_BYTE_ARRAY)));
}
m = new Put(row, currentValueKV.getTimestamp());
m.getFamilyMap().putAll(append.getFamilyMap());
break;
case DROP_SEQUENCE:
m = new Delete(row, clientTimestamp, null);
break;
case CREATE_SEQUENCE:
m = new Put(row, clientTimestamp);
m.getFamilyMap().putAll(append.getFamilyMap());
break;
}
if (!hadClientTimestamp) {
for (List<KeyValue> kvs : m.getFamilyMap().values()) {
for (KeyValue kv : kvs) {
kv.updateLatestStamp(clientTimestampBuf);
}
}
}
@SuppressWarnings("unchecked")
Pair<Mutation, Integer>[] mutations = new Pair[1];
mutations[0] = new Pair<Mutation, Integer>(m, lid);
region.batchMutate(mutations);
long serverTimestamp = MetaDataUtil.getClientTimeStamp(m);
// Return result with single KeyValue. The only piece of information
// the client cares about is the timestamp, which is the timestamp of
// when the mutation was actually performed (useful in the case of .
return new Result(
Collections.singletonList(
KeyValueUtil.newKeyValue(
row,
PhoenixDatabaseMetaData.SEQUENCE_FAMILY_BYTES,
QueryConstants.EMPTY_COLUMN_BYTES,
serverTimestamp,
SUCCESS_VALUE)));
} finally {
region.releaseRowLock(lid);
}
} catch (Throwable t) {
ServerUtil.throwIOException("Increment of sequence " + Bytes.toStringBinary(row), t);
return null; // Impossible
} finally {
region.closeRegionOperation();
}
}