UserScan的处理流程分析

UserScan的处理流程分析

 

前置说明

 

Userscan是通过client或cp中发起的scanner操作。

 

在Scan中通过caching属性来返回可以返回多少条数据，每次进行next时。

 

通过batch属性来设置每次在rs端每次next kv时，可读取多少个kv，(在同一行的情况下)

 

在生成Scan实例时，最好是把family与column都设置上，这样能保证查询的最高效.

 

client端通过生成Scan实例，通过HTable下的如下方法得到ClientScanner实例

 

public ResultScanner getScanner(final Scan scan)

 

在生成的ClientScanner实例中的callable属性的值为生成的一个ScannerCallable实例。

 

并通过callable.prepare(tries != 0);方法得到此scan的startkey所在的region的location.在meta表中。

 

把startkey对应的location中得到此location的HRegionInfo信息。

 

并设置ClientScanner.currentRegion的值为当前的region.也就是startkey所在的region.

 

 

 

通过ClientScanner.next向rs发起rpc调用操作。调用HRegionServer.scan

 

public ScanResponse scan(finalRpcControllercontroller, final ScanRequest request)

 

 

 

 

 

ClientScanner.next时，首先是发起openScanner操作，得到一个ScannerId

 

通过ScannerCallable.call方法：

 

if (scannerId == -1L) {

 

this.scannerId = openScanner();

 

} else {

 

openScanner方法:中发起一个scan操作，通过rpc调用rs.scan

 

ScanRequest request =

 

RequestConverter.buildScanRequest(

 

getLocation().getRegionInfo().getRegionName(),

 

this.scan, 0, false);

 

try {

 

ScanResponse response = getStub().scan(null, request);

 

longid = response.getScannerId();

 

if (logScannerActivity) {

 

LOG.info("Open scanner=" + id + " for scan=" + scan.toString()

 

+ " on region " + getLocation().toString());

 

}

 

returnid;

 

 

 

HregionServer.scan中对openScanner的处理：

 

public ScanResponse scan(finalRpcControllercontroller, final ScanRequest request)

 

throws ServiceException {

 

Leases.Lease lease = null;

 

String scannerName = null;

 

........................................很多代码没有显示

 

requestCount.increment();

 

 

 

intttl = 0;

 

HRegion region = null;

 

RegionScannerscanner = null;

 

RegionScannerHolder rsh = null;

 

booleanmoreResults = true;

 

booleancloseScanner = false;

 

ScanResponse.Builder builder = ScanResponse.newBuilder();

 

if (request.hasCloseScanner()) {

 

closeScanner = request.getCloseScanner();

 

}

 

introws = 1;

 

if (request.hasNumberOfRows()) {

 

rows = request.getNumberOfRows();

 

}

 

if (request.hasScannerId()) {

 

.................................很多代码没有显示

 

} else {

 

得到请求的HRegion实例,也就是startkey所在的HRegion

 

region = getRegion(request.getRegion());

 

ClientProtos.Scan protoScan = request.getScan();

 

booleanisLoadingCfsOnDemandSet = protoScan.hasLoadColumnFamiliesOnDemand();

 

Scan scan = ProtobufUtil.toScan(protoScan);

 

// if the request doesn't set this, get the default region setting.

 

if (!isLoadingCfsOnDemandSet) {

 

scan.setLoadColumnFamiliesOnDemand(region.isLoadingCfsOnDemandDefault());

 

}

 

scan.getAttribute(Scan.SCAN_ATTRIBUTES_METRICS_ENABLE);

 

如果scan没有设置family,把region中所有的family当成scan的family

 

region.prepareScanner(scan);

 

if (region.getCoprocessorHost() != null) {

 

scanner = region.getCoprocessorHost().preScannerOpen(scan);

 

}

 

if (scanner == null) {

 

执行HRegion.getScanner方法。生成HRegion.RegionScannerImpl方法

 

scanner = region.getScanner(scan);

 

}

 

if (region.getCoprocessorHost() != null) {

 

scanner = region.getCoprocessorHost().postScannerOpen(scan, scanner);

 

}

 

把生成的RegionScanner添加到scanners集合容器中。并设置scannerid(一个随机的值),

 

scannername是scannerid的string版本。添加过期监控处理，

 

通过hbase.client.scanner.timeout.period配置过期时间，默认值为60000ms

 

老版本通过hbase.regionserver.lease.period配置。

 

过期检查线程通过Leases完成。对scanner的过期处理通过一个

 

HregionServer.ScannerListener.leaseExpired实例来完成。

 

 

 

scannerId = addScanner(scanner, region);

 

scannerName = String.valueOf(scannerId);

 

ttl = this.scannerLeaseTimeoutPeriod;

 

}

 

............................................很多代码没有显示

 

 

 

Hregion.getScanner方法生成RegionScanner实例流程

 

 

 

publicRegionScannergetScanner(Scan scan) throws IOException {

 

returngetScanner(scan, null);

 

}

 

 

 

层次的调用,此时传入的kvscannerlist为null

 

protectedRegionScannergetScanner(Scan scan,

 

List<KeyValueScanner> additionalScanners) throws IOException {

 

startRegionOperation(Operation.SCAN);

 

try {

 

// Verify families are all valid

 

prepareScanner(scan);

 

if(scan.hasFamilies()) {

 

for(byte [] family : scan.getFamilyMap().keySet()) {

 

checkFamily(family);

 

}

 

}

 

returninstantiateRegionScanner(scan, additionalScanners);

 

} finally {

 

closeRegionOperation();

 

}

 

}

 

 

 

最终生成一个HRegion.RegionScannerImpl实例

 

protectedRegionScannerinstantiateRegionScanner(Scan scan,

 

List<KeyValueScanner> additionalScanners) throws IOException {

 

returnnewRegionScannerImpl(scan, additionalScanners, this);

 

}

 

 

 

RegionScanner实例的生成构造方法：

 

RegionScannerImpl(Scan scan, List<KeyValueScanner> additionalScanners, HRegion region)

 

throws IOException {

 

 

 

this.region = region;

 

this.maxResultSize = scan.getMaxResultSize();

 

if (scan.hasFilter()) {

 

this.filter = newFilterWrapper(scan.getFilter());

 

} else {

 

this.filter = null;

 

}

 

 

 

this.batch = scan.getBatch();

 

if (Bytes.equals(scan.getStopRow(), HConstants.EMPTY_END_ROW) && !scan.isGetScan()) {

 

this.stopRow = null;

 

} else {

 

this.stopRow = scan.getStopRow();

 

}

 

// If we are doing a get, we want to be [startRow,endRow] normally

 

// it is [startRow,endRow) and if startRow=endRow we get nothing.

 

this.isScan = scan.isGetScan() ? -1 : 0;

 

 

 

// synchronize on scannerReadPoints so that nobody calculates

 

// getSmallestReadPoint, before scannerReadPoints is updated.

 

IsolationLevelisolationLevel = scan.getIsolationLevel();

 

synchronized(scannerReadPoints) {

 

if (isolationLevel == IsolationLevel.READ_UNCOMMITTED) {

 

// This scan can read even uncommitted transactions

 

this.readPt = Long.MAX_VALUE;

 

MultiVersionConsistencyControl.setThreadReadPoint(this.readPt);

 

} else {

 

this.readPt = MultiVersionConsistencyControl.resetThreadReadPoint(mvcc);

 

}

 

scannerReadPoints.put(this, this.readPt);

 

}

 

 

 

// Here we separate all scanners into two lists - scanner that provide data required

 

// by the filter to operate (scanners list) and all others (joinedScanners list).

 

List<KeyValueScanner> scanners = newArrayList<KeyValueScanner>();

 

List<KeyValueScanner> joinedScanners = newArrayList<KeyValueScanner>();

 

if (additionalScanners != null) {

 

scanners.addAll(additionalScanners);

 

}

 

迭代每一个要进行scan的store。生成具体的StoreScanner实例。通常情况下joinedHead的值为null

 

for (Map.Entry<byte[], NavigableSet<byte[]>> entry :

 

scan.getFamilyMap().entrySet()) {

 

Storestore = stores.get(entry.getKey());

 

生成StoreScanner实例。通过HStore.getScanner(scan,columns);

 

KeyValueScannerscanner = store.getScanner(scan, entry.getValue());

 

if (this.filter == null || !scan.doLoadColumnFamiliesOnDemand()

 

|| this.filter.isFamilyEssential(entry.getKey())) {

 

scanners.add(scanner);

 

} else {

 

joinedScanners.add(scanner);

 

}

 

}

 

生成KeyValueHeap实例，把所有的storescanner的开始位置移动到startkey的位置并得到top的StoreScanner,

 

this.storeHeap = newKeyValueHeap(scanners, comparator);

 

if (!joinedScanners.isEmpty()) {

 

this.joinedHeap = newKeyValueHeap(joinedScanners, comparator);

 

}

 

}

 

 

 

得到StoreScanner实例的HStore.getScanner(scan,columns)方法

 

publicKeyValueScannergetScanner(Scan scan,

 

finalNavigableSet<byte []> targetCols) throws IOException {

 

lock.readLock().lock();

 

try {

 

KeyValueScannerscanner = null;

 

if (this.getCoprocessorHost() != null) {

 

scanner = this.getCoprocessorHost().preStoreScannerOpen(this, scan, targetCols);

 

}

 

if (scanner == null) {

 

scanner = newStoreScanner(this, getScanInfo(), scan, targetCols);

 

}

 

returnscanner;

 

} finally {

 

lock.readLock().unlock();

 

}

 

}

 

生成StoreScanner的构造方法：

 

publicStoreScanner(Storestore, ScanInfo scanInfo, Scan scan, finalNavigableSet<byte[]> columns)

 

throws IOException {

 

this(store, scan.getCacheBlocks(), scan, columns, scanInfo.getTtl(),

 

scanInfo.getMinVersions());

 

如果设置有scan的_raw_属性时，columns的值需要为null

 

if (columns != null && scan.isRaw()) {

 

thrownewDoNotRetryIOException(

 

"Cannot specify any column for a raw scan");

 

}

 

matcher = newScanQueryMatcher(scan, scanInfo, columns,

 

ScanType.USER_SCAN, Long.MAX_VALUE, HConstants.LATEST_TIMESTAMP,

 

oldestUnexpiredTS);

 

得到StoreFileScanner,StoreFileScanner中引用的StoreFile.Reader中引用HFileReaderV2,

 

HFileReaderV2的实例在StoreFile.Reader中如果已经存在，不会重新创建，这样会加快scanner的创建时间。

 

// Pass columns to try to filter out unnecessary StoreFiles.

 

List<KeyValueScanner> scanners = getScannersNoCompaction();

 

 

 

// Seek all scanners to the start of the Row (or if the exact matching row

 

// key does not exist, then to the start of the next matching Row).

 

// Always check bloom filter to optimize the top row seek for delete

 

// family marker.

 

if (explicitColumnQuery && lazySeekEnabledGlobally) {

 

for (KeyValueScannerscanner : scanners) {

 

scanner.requestSeek(matcher.getStartKey(), false, true);

 

}

 

} else {

 

if (!isParallelSeekEnabled) {

 

for (KeyValueScannerscanner : scanners) {

 

scanner.seek(matcher.getStartKey());

 

}

 

} else {

 

parallelSeek(scanners, matcher.getStartKey());

 

}

 

}

 

 

 

// set storeLimit

 

this.storeLimit = scan.getMaxResultsPerColumnFamily();

 

 

 

// set rowOffset

 

this.storeOffset = scan.getRowOffsetPerColumnFamily();

 

 

 

// Combine all seeked scanners with a heap

 

heap = newKeyValueHeap(scanners, store.getComparator());

 

注册，如果有storefile更新时，把更新后的storefile添加到这个StoreScanner中来。

 

this.store.addChangedReaderObserver(this);

 

}

 

 

 

发起scan的rpc操作

 

client端发起openScanner操作后，得到一个scannerId.此时发起scan操作。

 

通过ScannerCallable.call中发起call的操作，在scannerId不等于-1时，

 

 

 

Result [] rrs = null;

 

ScanRequest request = null;

 

try {

 

incRPCcallsMetrics();

 

request = RequestConverter.buildScanRequest(scannerId, caching, false, nextCallSeq);

 

ScanResponse response = null;

 

PayloadCarryingRpcController controller = newPayloadCarryingRpcController();

 

try {

 

controller.setPriority(getTableName());

 

response = getStub().scan(controller, request);

 

...................................此处省去一些代码

 

nextCallSeq++;

 

longtimestamp = System.currentTimeMillis();

 

// Results are returned via controller

 

CellScannercellScanner = controller.cellScanner();

 

rrs = ResponseConverter.getResults(cellScanner, response);

 

 

 

 

 

HregionServer.scan方法中对scan时的处理流程：

 

得到scan中的caching属性的值，此值主要用来响应client返回的条数。如果一行数据包含多个kv，算一条

 

introws = 1;

 

if (request.hasNumberOfRows()) {

 

rows = request.getNumberOfRows();

 

}

 

如果client传入的scannerId有值，也就是不等于-1时，表示不是openScanner操作，检查scannerid是否过期

 

if (request.hasScannerId()) {

 

rsh = scanners.get(scannerName);

 

if (rsh == null) {

 

LOG.info("Client tried to access missing scanner " + scannerName);

 

thrownewUnknownScannerException(

 

"Name: " + scannerName + ", already closed?");

 

}

 

此处主要是检查region是否发生过split操作。如果是会出现NotServingRegionException操作。

 

scanner = rsh.s;

 

HRegionInfo hri = scanner.getRegionInfo();

 

region = getRegion(hri.getRegionName());

 

if (region != rsh.r) { // Yes, should be the same instance

 

thrownewNotServingRegionException("Region was re-opened after the scanner"

 

+ scannerName + " was created: " + hri.getRegionNameAsString());

 

}

 

} else {

 

...................................此处省去一些生成Regionscanner的代码

 

}

 

表示有设置caching,如果是执行scan,此时的默认值为1,当前scan中设置有caching后，使用scan中设置的值

 

if (rows > 0) {

 

// if nextCallSeq does not match throw Exception straight away. This needs to be

 

// performed even before checking of Lease.

 

// See HBASE-5974

 

是否有配置nextCallSeq的值，第一次调用时，此值为0,每调用一次加一,client也一样，每调用一次加一。

 

if (request.hasNextCallSeq()) {

 

if (rsh == null) {

 

rsh = scanners.get(scannerName);

 

}

 

if (rsh != null) {

 

if (request.getNextCallSeq() != rsh.nextCallSeq) {

 

thrownewOutOfOrderScannerNextException("Expected nextCallSeq: " + rsh.nextCallSeq

 

+ " But the nextCallSeq got from client: " + request.getNextCallSeq() +

 

"; request=" + TextFormat.shortDebugString(request));

 

}

 

// Increment the nextCallSeq value which is the next expected from client.

 

rsh.nextCallSeq++;

 

}

 

}

 

try {

 

先从租约管理中移出此租约，防止查找时间大于过期时间而出现的超时

 

// Remove lease while its being processed in server; protects against case

 

// where processing of request takes > lease expiration time.

 

lease = leases.removeLease(scannerName);

 

生成要返回的条数的一个列表，scan.caching

 

List<Result> results = newArrayList<Result>(rows);

 

longcurrentScanResultSize = 0;

 

 

 

booleandone = false;

 

调用cp的preScannernext,如果返回为true,表示不在执行scan操作。

 

// Call coprocessor. Get region info from scanner.

 

if (region != null && region.getCoprocessorHost() != null) {

 

Boolean bypass = region.getCoprocessorHost().preScannerNext(

 

scanner, results, rows);

 

if (!results.isEmpty()) {

 

for (Result r : results) {

 

if (maxScannerResultSize < Long.MAX_VALUE){

 

for (Cellkv : r.rawCells()) {

 

// TODO

 

currentScanResultSize += KeyValueUtil.ensureKeyValue(kv).heapSize();

 

}

 

}

 

}

 

}

 

if (bypass != null && bypass.booleanValue()) {

 

done = true;

 

}

 

}

 

执行scan操作。Cp的preScannerNext返回为false,或没有设置cp(主要是RegionObServer)

 

返回给client的最大size通过hbase.client.scanner.max.result.size配置，默认为long.maxvalue

 

如果scan也设置有maxResultSize,使用scan设置的值

 

if (!done) {

 

longmaxResultSize = scanner.getMaxResultSize();

 

if (maxResultSize <= 0) {

 

maxResultSize = maxScannerResultSize;

 

}

 

List<Cell> values = newArrayList<Cell>();

 

MultiVersionConsistencyControl.setThreadReadPoint(scanner.getMvccReadPoint());

 

region.startRegionOperation(Operation.SCAN);

 

try {

 

inti = 0;

 

synchronized(scanner) {

 

此处开始迭代，开始调用regionScanner(HRegion.RegionScannerImpl.nextRaw(List))进行查找，

 

迭代的长度为scan设置的caching的大小,如果执行RegionScanner.nextRaw(List)返回为false,时也会停止迭代

 

for (; i < rows

 

&& currentScanResultSize < maxResultSize; i++) {

 

返回的true表示还有数据，可以接着查询，否则表示此region中已经没有符合条件的数据了。

 

// Collect values to be returned here

 

booleanmoreRows = scanner.nextRaw(values);

 

if (!values.isEmpty()) {

 

if (maxScannerResultSize < Long.MAX_VALUE){

 

for (Cellkv : values) {

 

currentScanResultSize += KeyValueUtil.ensureKeyValue(kv).heapSize();

 

}

 

}

 

results.add(Result.create(values));

 

}

 

if (!moreRows) {

 

break;

 

}

 

values.clear();

 

}

 

}

 

region.readRequestsCount.add(i);

 

} finally {

 

region.closeRegionOperation();

 

}

 

 

 

// coprocessor postNext hook

 

if (region != null && region.getCoprocessorHost() != null) {

 

region.getCoprocessorHost().postScannerNext(scanner, results, rows, true);

 

}

 

}

 

如果没有可以再查找的数据时，设置response的moreResults为false

 

// If the scanner's filter - if any - is done with the scan

 

// and wants to tell the client to stop the scan. This is done by passing

 

// a null result, and setting moreResults to false.

 

if (scanner.isFilterDone() && results.isEmpty()) {

 

moreResults = false;

 

results = null;

 

} else {

 

添加结果到response中，如果hbase.client.rpc.codec配置有codec的值，

 

默认取hbase.client.default.rpc.codec配置的值，默认为KeyValueCodec

 

如果上面说的codec配置不为null时，把results生成为一个iterator,并生成一个匿名的CallScanner实现类

 

设置到scan时传入的controller中。这样能提升查询数据的读取性能。

 

如果没有配置codec时，默认直接把results列表设置到response中，这样响应的数据可能会比较大。

 

addResults(builder, results, controller);

 

}

 

} finally {

 

重新把租约放入到租约检查管理器中，此租约主要来检查client多长时间没有发起过scan的操作。

 

// We're done. On way out re-add the above removed lease.

 

// Adding resets expiration time on lease.

 

if (scanners.containsKey(scannerName)) {

 

if (lease != null) leases.addLease(lease);

 

ttl = this.scannerLeaseTimeoutPeriod;

 

}

 

}

 

}

 

 

 

client端获取响应的数据：ScannerCallable.call方法中

 

rrs = ResponseConverter.getResults(cellScanner, response);

 

 

 

ResponseConverter.getResults方法的实现

 

publicstatic Result[] getResults(CellScannercellScanner, ScanResponse response)

 

throws IOException {

 

if (response == null) returnnull;

 

// If cellscanner, then the number of Results to return is the count of elements in the

 

// cellsPerResult list. Otherwise, it is how many results are embedded inside the response.

 

intnoOfResults = cellScanner != null?

 

response.getCellsPerResultCount(): response.getResultsCount();

 

Result[] results = new Result[noOfResults];

 

for (inti = 0; i < noOfResults; i++) {

 

cellScanner如果codec配置为有值时，在rs响应时会生成一个匿名的实现

 

if (cellScanner != null) {

 

......................................

 

intnoOfCells = response.getCellsPerResult(i);

 

List<Cell> cells = newArrayList<Cell>(noOfCells);

 

for (intj = 0; j < noOfCells; j++) {

 

try {

 

if (cellScanner.advance() == false) {

 

.....................................

 

String msg = "Results sent from server=" + noOfResults + ". But only got " + i

 

+ " results completely at client. Resetting the scanner to scan again.";

 

LOG.error(msg);

 

thrownewDoNotRetryIOException(msg);

 

}

 

} catch (IOException ioe) {

 

...........................................

 

LOG.error("Exception while reading cells from result."

 

+ "Resetting the scanner to scan again.", ioe);

 

thrownewDoNotRetryIOException("Resetting the scanner.", ioe);

 

}

 

cells.add(cellScanner.current());

 

}

 

results[i] = Result.create(cells);

 

} else {

 

否则，没有设置codec，直接从response中读取出来数据，

 

// Result is pure pb.

 

results[i] = ProtobufUtil.toResult(response.getResults(i));

 

}

 

}

 

returnresults;

 

}

 

 

 

在ClientScanner.next方法中，如果还没有达到scan的caching的值，(默认为1)也就是countdown的值还不等于0

 

,countdown的值为得到一个Result时减1，通过nextScanner重新得到下一个region，并发起连接去scan数据。

 

 

 

Do{

 

.........................此处省去一些代码。

 

if (values != null && values.length > 0) {

 

for (Result rs : values) {

 

cache.add(rs);

 

for (Cellkv : rs.rawCells()) {

 

// TODO make method in Cell or CellUtil

 

remainingResultSize -= KeyValueUtil.ensureKeyValue(kv).heapSize();

 

}

 

countdown--;

 

this.lastResult = rs;

 

}

 

}

 

} while (remainingResultSize > 0 && countdown > 0 && nextScanner(countdown, values == null));

 

 

 

对于这种类型的查询操作，可以使用得到一个ClientScanner后，不执行close操作。

 

在rs的timeout前每次定期去从rs中拿一定量的数据下来。缓存到ClientScanner的cache中。

 

每次next时从cache中直接拿数据

 

 

 

Hregion.RegionScannerImpl.nextRaw(list)方法分析

 

RegionScannerImpl是对RegionScanner接口的实现。

 

Rs的scan在执行时通过regionScanner.nextRaw(list)来获取数据。

 

通过regionScanner.isFilterDone来检查此region的查找是否完成。

 

 

 

调用nextRaw方法，此方法调用另一个重载方法，batch是scan中设置的每次可查询最大的单行中的多少个kv的kv个数

 

publicbooleannextRaw(List<Cell> outResults)

 

throws IOException {

 

returnnextRaw(outResults, batch);

 

}

 

 

 

publicbooleannextRaw(List<Cell> outResults, intlimit) throws IOException {

 

booleanreturnResult;

 

调用nextInternal方法。

 

if (outResults.isEmpty()) {

 

// Usually outResults is empty. This is true when next is called

 

// to handle scan or get operation.

 

returnResult = nextInternal(outResults, limit);

 

} else {

 

List<Cell> tmpList = newArrayList<Cell>();

 

returnResult = nextInternal(tmpList, limit);

 

outResults.addAll(tmpList);

 

}

 

调用filter.reset方法，清空当前row的filter的相关信息。

 

ResetFilters();

 

如果filter.filterAllRemaining()的返回值为true,时表示当前region的查找条件已经结束，不能在执行查找操作。

 

没有可以接着查找的需要，也就是没有更多要查找的行了。

 

if (isFilterDone()) {

 

returnfalse;

 

}

 

................................此处省去一些代码

 

returnreturnResult;

 

}

 

 

 

nextInternal方法处理流程：

 

privatebooleannextInternal(List<Cell> results, intlimit)

 

throws IOException {

 

if (!results.isEmpty()) {

 

thrownewIllegalArgumentException("First parameter should be an empty list");

 

}

 

RpcCallContextrpcCall = RpcServer.getCurrentCall();

 

// The loop here is used only when at some point during the next we determine

 

// that due to effects of filters or otherwise, we have an empty row in the result.

 

// Then we loop and try again. Otherwise, we must get out on the first iteration via return,

 

// "true" if there's more data to read, "false" if there isn't (storeHeap is at a stop row,

 

// and joinedHeap has no more data to read for the last row (if set, joinedContinuationRow).

 

while (true) {

 

if (rpcCall != null) {

 

// If a user specifies a too-restrictive or too-slow scanner, the

 

// client might time out and disconnect while the server side

 

// is still processing the request. We should abort aggressively

 

// in that case.

 

longafterTime = rpcCall.disconnectSince();

 

if (afterTime >= 0) {

 

thrownewCallerDisconnectedException(

 

"Aborting on region " + getRegionNameAsString() + ", call " +

 

this + " after " + afterTime + " ms, since " +

 

"caller disconnected");

 

}

 

}

 

得到通过startkey seek后当前最小的一个kv。

 

// Let's see what we have in the storeHeap.

 

KeyValue current = this.storeHeap.peek();

 

 

 

byte[] currentRow = null;

 

intoffset = 0;

 

shortlength = 0;

 

if (current != null) {

 

currentRow = current.getBuffer();

 

offset = current.getRowOffset();

 

length = current.getRowLength();

 

}

 

检查是否到了stopkey,如果是，返回false,joinedContinuationRow是多个cf的关联查找，不用去管它

 

booleanstopRow = isStopRow(currentRow, offset, length);

 

// Check if we were getting data from the joinedHeap and hit the limit.

 

// If not, then it's main path - getting results from storeHeap.

 

if (joinedContinuationRow == null) {

 

// First, check if we are at a stop row. If so, there are no more results.

 

if (stopRow) {

 

如果是stopRow,同时filter.hasFilterRow返回为true时，

 

可通过filterRowCells来检查要返回的kvlist,也可以用来修改要返回的kvlist

 

if (filter != null && filter.hasFilterRow()) {

 

filter.filterRowCells(results);

 

}

 

returnfalse;

 

}

 

通过filter.filterRowkey来过滤检查key是否需要排除，如果是排除返回true,否则返回false

 

// Check if rowkey filter wants to exclude this row. If so, loop to next.

 

// Technically, if we hit limits before on this row, we don't need this call.

 

if (filterRowKey(currentRow, offset, length)) {

 

如果rowkey是需要排除的rowkey,检查是否有下一行数据。如果没有下一行数据，返回flase,表示当前region查找结束

 

否则清空当前的results,重新进行查找

 

booleanmoreRows = nextRow(currentRow, offset, length);

 

if (!moreRows) returnfalse;

 

results.clear();

 

continue;

 

}

 

开始执行region下此scan需要的所有store的StoreScanner的next进行查找，把查找的结果放到results列表中。

 

如果一行中包含有多个kv，现在查找这些kv达到传入的limit的大小的时候，返回kv_limit的一个空的kv。

 

(查找的大小已经达到limit(batch)的一行最大scan的kv个数，返回kv_limit),

 

否则表示还没有查找到limit的kv个数，但是当前row对应的所有达到条件的kv都已经查找完成，返回最后一个kv。

 

返回的kv如果不是kv_limit，那么有可能是null或者是下一行的第一个kv.

 

KeyValue nextKv = populateResult(results, this.storeHeap, limit, currentRow, offset,

 

length);

 

如果达到limit的限制时，filter.hasFilterRow的值一定得是false,

 

否则会throw IncompatibleFilterException

 

如果达到limit的限制时，返回true,当前row的所有kv查找结束，返回true可以接着向下查找

 

提示：如果hbase一行数据中可能包含多个kv时，最好是在scan时设置batch的属性，否则会一直查找到所有的kv结束

 

// Ok, we are good, let's try to get some results from the main heap.

 

if (nextKv == KV_LIMIT) {

 

if (this.filter != null && filter.hasFilterRow()) {

 

thrownewIncompatibleFilterException(

 

"Filter whose hasFilterRow() returns true is incompatible with scan with limit!");

 

}

 

returntrue; // We hit the limit.

 

}

 

是否到结束行，从这一行代码中可以看出，stoprow是不包含的，因为nextKv肯定是下一行row中第一个kv的值

 

stopRow = nextKv == null ||

 

isStopRow(nextKv.getBuffer(), nextKv.getRowOffset(), nextKv.getRowLength());

 

// save that the row was empty before filters applied to it.

 

finalbooleanisEmptyRow = results.isEmpty();

 

 

 

如果是stopRow,同时filter.hasFilterRow返回为true时，

 

可通过filterRowCells来检查要返回的kvlist,也可以用来修改要返回的kvlist

 

// We have the part of the row necessary for filtering (all of it, usually).

 

// First filter with the filterRow(List).

 

if (filter != null && filter.hasFilterRow()) {

 

filter.filterRowCells(results);

 

}

 

如果当前row的查找没有找到合法的kv，也就是results的列表没有值，检查是否还有下一行，

 

如果有，重新进行查找，否则表示当前region的查找最结尾处，不能再进行查找，返回fasle

 

if (isEmptyRow) {

 

booleanmoreRows = nextRow(currentRow, offset, length);

 

if (!moreRows) returnfalse;

 

results.clear();

 

// This row was totally filtered out, if this is NOT the last row,

 

// we should continue on. Otherwise, nothing else to do.

 

if (!stopRow) continue;

 

returnfalse;

 

}

 

 

 

// Ok, we are done with storeHeap for this row.

 

// Now we may need to fetch additional, non-essential data into row.

 

// These values are not needed for filter to work, so we postpone their

 

// fetch to (possibly) reduce amount of data loads from disk.

 

if (this.joinedHeap != null) {

 

..................................进行关联查找的代码，不显示，也不分析

 

}

 

} else {

 

多个store进行关联查询，不分析，通常情况不会有

 

// Populating from the joined heap was stopped by limits, populate some more.

 

populateFromJoinedHeap(results, limit);

 

}

 

 

 

// We may have just called populateFromJoinedMap and hit the limits. If that is

 

// the case, we need to call it again on the next next() invocation.

 

if (joinedContinuationRow != null) {

 

returntrue;

 

}

 

如果这次的查找,results的结果为空，表示没有查找到结果，检查是否还有下一行数据，如果有重新进行查找，

 

否则返回false表示此region的查找结束

 

// Finally, we are done with both joinedHeap and storeHeap.

 

// Double check to prevent empty rows from appearing in result. It could be

 

// the case when SingleColumnValueExcludeFilter is used.

 

if (results.isEmpty()) {

 

booleanmoreRows = nextRow(currentRow, offset, length);

 

if (!moreRows) returnfalse;

 

if (!stopRow) continue;

 

}

 

非stoprow时，表示还可以有下一行的数据，也就是可以接着进行next操作。否则表示此region的查找结束

 

// We are done. Return the result.

 

return !stopRow;

 

}

 

}

 

 

 

UserScan时的ScanQueryMatcher.match方法处理

 

user scan时的ScanQueryMatcher在newRegionScannerImpl(scan, additionalScanners, this);时生成。

 

在生成StoreScanner时通过如下代码生成matcher实例。

 

 

 

matcher = newScanQueryMatcher(scan, scanInfo, columns,

 

ScanType.USER_SCAN, Long.MAX_VALUE, HConstants.LATEST_TIMESTAMP,

 

oldestUnexpiredTS);

 

 

 

matcher.isUserScan的值此时为true.

 

 

 

publicMatchCodematch(KeyValue kv) throws IOException {

 

检查当前region的查找是否结束。pageFilter就是通过控制此filter中的方法来检查是否需要

 

if (filter != null && filter.filterAllRemaining()) {

 

returnMatchCode.DONE_SCAN;

 

}

 

 

 

byte [] bytes = kv.getBuffer();

 

intoffset = kv.getOffset();

 

 

 

intkeyLength = Bytes.toInt(bytes, offset, Bytes.SIZEOF_INT);

 

offset += KeyValue.ROW_OFFSET;

 

 

 

intinitialOffset = offset;

 

 

 

shortrowLength = Bytes.toShort(bytes, offset, Bytes.SIZEOF_SHORT);

 

offset += Bytes.SIZEOF_SHORT;

 

检查传入的kv是否是当前行的kv，也就是rowkey是否相同，如果当前的rowkey小于传入的rowkey，

 

表示现在已经next到下一行了，返回DONE,表示当前行查找结束

 

intret = this.rowComparator.compareRows(row, this.rowOffset, this.rowLength,

 

bytes, offset, rowLength);

 

if (ret <= -1) {

 

returnMatchCode.DONE;

 

} elseif (ret >= 1) {

 

如果当前的rowkey大于传入的rowkey，表示当前next出来的kv比现在的kv要小，执行nextrow操作。

 

// could optimize this, if necessary?

 

// Could also be called SEEK_TO_CURRENT_ROW, but this

 

// should be rare/never happens.

 

returnMatchCode.SEEK_NEXT_ROW;

 

}

 

是否跳过当前行的其它kv比较，这是一个优化项。

 

// optimize case.

 

if (this.stickyNextRow)

 

returnMatchCode.SEEK_NEXT_ROW;

 

如果当前行的所有要查找的(scan)column都查找完成了，其它的当前行中非要scan的kv，

 

直接不比较,执行nextrow操作。

 

if (this.columns.done()) {

 

stickyNextRow = true;

 

returnMatchCode.SEEK_NEXT_ROW;

 

}

 

 

 

//Passing rowLength

 

offset += rowLength;

 

 

 

//Skipping family

 

bytefamilyLength = bytes [offset];

 

offset += familyLength + 1;

 

 

 

intqualLength = keyLength -

 

(offset - initialOffset) - KeyValue.TIMESTAMP_TYPE_SIZE;

 

检查当前KV的TTL是否过期，如果过期，检查是否SCAN中还有下一个COLUMN，如果有返回SEEK_NEXT_COL。

 

否则返回SEEK_NEXT_ROW。

 

longtimestamp = Bytes.toLong(bytes, initialOffset + keyLength - KeyValue.TIMESTAMP_TYPE_SIZE);

 

// check for early out based on timestamp alone

 

if (columns.isDone(timestamp)) {

 

returncolumns.getNextRowOrNextColumn(bytes, offset, qualLength);

 

}

 

 

 

/*

 

* The delete logic is pretty complicated now.

 

* This is corroborated by the following:

 

* 1. The store might be instructed to keep deleted rows around.

 

* 2. A scan can optionally see past a delete marker now.

 

* 3. If deleted rows are kept, we have to find out when we can

 

* remove the delete markers.

 

* 4. Family delete markers are always first (regardless of their TS)

 

* 5. Delete markers should not be counted as version

 

* 6. Delete markers affect puts of the *same* TS

 

* 7. Delete marker need to be version counted together with puts

 

* they affect

 

*/

 

bytetype = bytes[initialOffset + keyLength – 1];

 

如果当前KV是删除的KV。

 

if (kv.isDelete()) {

 

此处会进入。把删除的KV添加到DeleteTracker中，默认是ScanDeleteTracker

 

if (!keepDeletedCells) {

 

// first ignore delete markers if the scanner can do so, and the

 

// range does not include the marker

 

//

 

// during flushes and compactions also ignore delete markers newer

 

// than the readpoint of any open scanner, this prevents deleted

 

// rows that could still be seen by a scanner from being collected

 

booleanincludeDeleteMarker = seePastDeleteMarkers ?

 

tr.withinTimeRange(timestamp) :

 

tr.withinOrAfterTimeRange(timestamp);

 

if (includeDeleteMarker

 

&& kv.getMvccVersion() <= maxReadPointToTrackVersions) {

 

this.deletes.add(bytes, offset, qualLength, timestamp, type);

 

}

 

// Can't early out now, because DelFam come before any other keys

 

}

 

此处的检查不会进入,userscan不保留删除的数据

 

if (retainDeletesInOutput

 

|| (!isUserScan && (EnvironmentEdgeManager.currentTimeMillis() - timestamp) <= timeToPurgeDeletes)

 

|| kv.getMvccVersion() > maxReadPointToTrackVersions) {

 

// always include or it is not time yet to check whether it is OK

 

// to purge deltes or not

 

if (!isUserScan) {

 

// if this is not a user scan (compaction), we can filter this deletemarker right here

 

// otherwise (i.e. a "raw" scan) we fall through to normal version and timerange checking

 

returnMatchCode.INCLUDE;

 

}

 

} elseif (keepDeletedCells) {

 

if (timestamp < earliestPutTs) {

 

// keeping delete rows, but there are no puts older than

 

// this delete in the store files.

 

returncolumns.getNextRowOrNextColumn(bytes, offset, qualLength);

 

}

 

// else: fall through and do version counting on the

 

// delete markers

 

} else {

 

returnMatchCode.SKIP;

 

}

 

// note the following next else if...

 

// delete marker are not subject to other delete markers

 

} elseif (!this.deletes.isEmpty()) {

 

如果deleteTracker中不为空时，也就是当前行中有删除的KV，检查当前KV是否是删除的KV

 

提示：删除的KV在compare时，比正常的KV要小，所以在执行next操作时，delete的KV会先被查找出来。

 

如果是删除的KV，根据KV的删除类型，如果是版本被删除，返回SKIP。

 

否则如果SCAN中还有下一个要SCAN的column时，返回SEEK_NEXT_COL。

 

否则表示当前行没有需要在进行查找的KV，返回SEEK_NEXT_ROW。

 

DeleteResultdeleteResult = deletes.isDeleted(bytes, offset, qualLength,

 

timestamp);

 

switch (deleteResult) {

 

caseFAMILY_DELETED:

 

caseCOLUMN_DELETED:

 

returncolumns.getNextRowOrNextColumn(bytes, offset, qualLength);

 

caseVERSION_DELETED:

 

caseFAMILY_VERSION_DELETED:

 

returnMatchCode.SKIP;

 

caseNOT_DELETED:

 

break;

 

default:

 

thrownewRuntimeException("UNEXPECTED");

 

}

 

}

 

检查KV的时间是否在SCAN要查找的时间范围内，

 

inttimestampComparison = tr.compare(timestamp);

 

如果大于SCAN的最大时间，返回SKIP。

 

if (timestampComparison >= 1) {

 

returnMatchCode.SKIP;

 

} elseif (timestampComparison <= -1) {

 

如果小于SCAN的最小时间，如果SCAN中还有下一个要SCAN的column时，返回SEEK_NEXT_COL。

 

否则表示当前行没有需要在进行查找的KV，返回SEEK_NEXT_ROW。

 

returncolumns.getNextRowOrNextColumn(bytes, offset, qualLength);

 

}

 

检查当前KV的column是否是SCAN中指定的column列表中包含的值，如果是INCLUDE。

 

否则如果SCAN中还有下一个要SCAN的column时，返回SEEK_NEXT_COL。

 

否则表示当前行没有需要在进行查找的KV，返回SEEK_NEXT_ROW。

 

// STEP 1: Check if the column is part of the requested columns

 

MatchCodecolChecker = columns.checkColumn(bytes, offset, qualLength, type);

 

如果column是SCAN中要查找的column之一

 

if (colChecker == MatchCode.INCLUDE) {

 

ReturnCodefilterResponse = ReturnCode.SKIP;

 

// STEP 2: Yes, the column is part of the requested columns. Check if filter is present

 

if (filter != null) {

 

执行filter.filterKeyValue操作。并返回filter过滤的结果

 

// STEP 3: Filter the key value and return if it filters out

 

filterResponse = filter.filterKeyValue(kv);

 

switch (filterResponse) {

 

caseSKIP:

 

returnMatchCode.SKIP;

 

caseNEXT_COL:

 

如果SCAN中还有下一个要SCAN的column时，返回SEEK_NEXT_COL。

 

否则表示当前行没有需要在进行查找的KV，返回SEEK_NEXT_ROW。

 

returncolumns.getNextRowOrNextColumn(bytes, offset, qualLength);

 

caseNEXT_ROW:

 

stickyNextRow = true;

 

returnMatchCode.SEEK_NEXT_ROW;

 

caseSEEK_NEXT_USING_HINT:

 

returnMatchCode.SEEK_NEXT_USING_HINT;

 

default:

 

//It means it is either include or include and seek next

 

break;

 

}

 

}

 

/*

 

* STEP 4: Reaching this step means the column is part of the requested columns and either

 

* the filter is null or the filter has returned INCLUDE or INCLUDE_AND_NEXT_COL response.

 

* Now check the number of versions needed. This method call returns SKIP, INCLUDE,

 

* INCLUDE_AND_SEEK_NEXT_ROW, INCLUDE_AND_SEEK_NEXT_COL.

 

*

 

* FilterResponse ColumnChecker Desired behavior

 

* INCLUDE SKIP row has already been included, SKIP.

 

* INCLUDE INCLUDE INCLUDE

 

* INCLUDE INCLUDE_AND_SEEK_NEXT_COL INCLUDE_AND_SEEK_NEXT_COL

 

* INCLUDE INCLUDE_AND_SEEK_NEXT_ROW INCLUDE_AND_SEEK_NEXT_ROW

 

* INCLUDE_AND_SEEK_NEXT_COL SKIP row has already been included, SKIP.

 

* INCLUDE_AND_SEEK_NEXT_COL INCLUDE INCLUDE_AND_SEEK_NEXT_COL

 

* INCLUDE_AND_SEEK_NEXT_COL INCLUDE_AND_SEEK_NEXT_COL INCLUDE_AND_SEEK_NEXT_COL

 

* INCLUDE_AND_SEEK_NEXT_COL INCLUDE_AND_SEEK_NEXT_ROW INCLUDE_AND_SEEK_NEXT_ROW

 

*

 

* In all the above scenarios, we return the column checker return value except for

 

* FilterResponse (INCLUDE_AND_SEEK_NEXT_COL) and ColumnChecker(INCLUDE)

 

*/

 

 

 

此处主要是检查KV的是否是SCAN的最大版本内，到这个地方，除非是KV超过了要SCAN的最大版本，或者KV的TTL过期。

 

否则肯定是会包含此KV的值。

 

 

 

colChecker =

 

columns.checkVersions(bytes, offset, qualLength, timestamp, type,

 

kv.getMvccVersion() > maxReadPointToTrackVersions);

 

//Optimize with stickyNextRow

 

stickyNextRow = colChecker == MatchCode.INCLUDE_AND_SEEK_NEXT_ROW ? true : stickyNextRow;

 

return (filterResponse == ReturnCode.INCLUDE_AND_NEXT_COL &&

 

colChecker == MatchCode.INCLUDE) ? MatchCode.INCLUDE_AND_SEEK_NEXT_COL

 

: colChecker;

 

}

 

stickyNextRow = (colChecker == MatchCode.SEEK_NEXT_ROW) ? true

 

: stickyNextRow;

 

returncolChecker;

 

}