Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HDFS-6355. Fix divide-by-zero, improper use of wall-clock time in BlockPoolSliceScanner (cmccabe) 

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/branch-2@1594339 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Colin McCabe 2014-05-13 19:25:23 +00:00
parent 96d4036449
commit 7b27dcf4dc
3 changed files with 26 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
View File

@ -214,6 +214,9 @@ Release 2.5.0 - UNRELEASED
HDFS-6305. WebHdfs response decoding may throw RuntimeExceptions (Daryn
Sharp via jeagles)
HDFS-6355. Fix divide-by-zero, improper use of wall-clock time in
BlockPoolSliceScanner (cmccabe)
Release 2.4.1 - UNRELEASED
INCOMPATIBLE CHANGES

27
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/BlockPoolSliceScanner.java
View File

@ -97,7 +97,7 @@ class BlockPoolSliceScanner {
private long totalTransientErrors = 0;
private final AtomicInteger totalBlocksScannedInLastRun = new AtomicInteger(); // Used for test only
private long currentPeriodStart = Time.now();
private long currentPeriodStart = Time.monotonicNow();
private long bytesLeft = 0; // Bytes to scan in this period
private long totalBytesToScan = 0;
private boolean isNewPeriod = true;
@ -260,7 +260,7 @@ class BlockPoolSliceScanner {
long period = Math.min(scanPeriod,
Math.max(blockMap.size(),1) * 600 * 1000L);
int periodInt = Math.abs((int)period);
return Time.now() - scanPeriod +
return Time.monotonicNow() - scanPeriod +
DFSUtil.getRandom().nextInt(periodInt);
}
@ -322,7 +322,7 @@ class BlockPoolSliceScanner {
info = new BlockScanInfo(block);
}
long now = Time.now();
long now = Time.monotonicNow();
info.lastScanType = type;
info.lastScanTime = now;
info.lastScanOk = scanOk;
@ -399,8 +399,9 @@ class BlockPoolSliceScanner {
}
private synchronized void adjustThrottler() {
long timeLeft = currentPeriodStart+scanPeriod - Time.now();
long bw = Math.max(bytesLeft*1000/timeLeft, MIN_SCAN_RATE);
long timeLeft = Math.max(1L,
currentPeriodStart + scanPeriod - Time.monotonicNow());
long bw = Math.max((bytesLeft * 1000) / timeLeft, MIN_SCAN_RATE);
throttler.setBandwidth(Math.min(bw, MAX_SCAN_RATE));
}
@ -523,7 +524,7 @@ class BlockPoolSliceScanner {
private boolean assignInitialVerificationTimes() {
//First updates the last verification times from the log file.
if (verificationLog != null) {
long now = Time.now();
long now = Time.monotonicNow();
RollingLogs.LineIterator logIterator = null;
try {
logIterator = verificationLog.logs.iterator(false);
@ -574,7 +575,7 @@ class BlockPoolSliceScanner {
// Initially spread the block reads over half of scan period
// so that we don't keep scanning the blocks too quickly when restarted.
long verifyInterval = Math.min(scanPeriod/(2L * numBlocks), 10*60*1000L);
long lastScanTime = Time.now() - scanPeriod;
long lastScanTime = Time.monotonicNow() - scanPeriod;
if (!blockInfoSet.isEmpty()) {
BlockScanInfo info;
@ -601,16 +602,16 @@ class BlockPoolSliceScanner {
// reset the byte counts :
bytesLeft = totalBytesToScan;
currentPeriodStart = Time.now();
currentPeriodStart = Time.monotonicNow();
isNewPeriod = true;
}
private synchronized boolean workRemainingInCurrentPeriod() {
if (bytesLeft <= 0 && Time.now() < currentPeriodStart + scanPeriod) {
if (bytesLeft <= 0 && Time.monotonicNow() < currentPeriodStart + scanPeriod) {
if (LOG.isDebugEnabled()) {
LOG.debug("Skipping scan since bytesLeft=" + bytesLeft + ", Start=" +
currentPeriodStart + ", period=" + scanPeriod + ", now=" +
Time.now() + " " + blockPoolId);
Time.monotonicNow() + " " + blockPoolId);
}
return false;
} else {
@ -633,7 +634,7 @@ class BlockPoolSliceScanner {
scan();
} finally {
totalBlocksScannedInLastRun.set(processedBlocks.size());
lastScanTime.set(Time.now());
lastScanTime.set(Time.monotonicNow());
}
}
@ -656,7 +657,7 @@ class BlockPoolSliceScanner {
while (datanode.shouldRun
&& !datanode.blockScanner.blockScannerThread.isInterrupted()
&& datanode.isBPServiceAlive(blockPoolId)) {
long now = Time.now();
long now = Time.monotonicNow();
synchronized (this) {
if ( now >= (currentPeriodStart + scanPeriod)) {
startNewPeriod();
@ -714,7 +715,7 @@ class BlockPoolSliceScanner {
int total = blockInfoSet.size();
long now = Time.now();
long now = Time.monotonicNow();
Date date = new Date();

18
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDatanodeBlockScanner.java
View File

@ -87,15 +87,15 @@ public class TestDatanodeBlockScanner {
throws IOException, TimeoutException {
URL url = new URL("http://localhost:" + infoPort +
"/blockScannerReport?listblocks");
long lastWarnTime = Time.now();
long lastWarnTime = Time.monotonicNow();
if (newTime <= 0) newTime = 1L;
long verificationTime = 0;
String block = DFSTestUtil.getFirstBlock(fs, file).getBlockName();
long failtime = (timeout <= 0) ? Long.MAX_VALUE
: Time.now() + timeout;
: Time.monotonicNow() + timeout;
while (verificationTime < newTime) {
if (failtime < Time.now()) {
if (failtime < Time.monotonicNow()) {
throw new TimeoutException("failed to achieve block verification after "
+ timeout + " msec. Current verification timestamp = "
+ verificationTime + ", requested verification time > "
@ -118,7 +118,7 @@ public class TestDatanodeBlockScanner {
}
if (verificationTime < newTime) {
long now = Time.now();
long now = Time.monotonicNow();
if ((now - lastWarnTime) >= 5*1000) {
LOG.info("Waiting for verification of " + block);
lastWarnTime = now;
@ -134,7 +134,7 @@ public class TestDatanodeBlockScanner {
@Test
public void testDatanodeBlockScanner() throws IOException, TimeoutException {
long startTime = Time.now();
long startTime = Time.monotonicNow();
Configuration conf = new HdfsConfiguration();
MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf).build();
@ -344,7 +344,7 @@ public class TestDatanodeBlockScanner {
conf.setLong(DFSConfigKeys.DFS_HEARTBEAT_INTERVAL_KEY, 3L);
conf.setBoolean(DFSConfigKeys.DFS_NAMENODE_REPLICATION_CONSIDERLOAD_KEY, false);
long startTime = Time.now();
long startTime = Time.monotonicNow();
MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf)
.numDataNodes(REPLICATION_FACTOR)
.build();
@ -428,10 +428,10 @@ public class TestDatanodeBlockScanner {
private static void waitForBlockDeleted(ExtendedBlock blk, int dnIndex,
long timeout) throws TimeoutException, InterruptedException {
File blockFile = MiniDFSCluster.getBlockFile(dnIndex, blk);
long failtime = Time.now()
long failtime = Time.monotonicNow()
+ ((timeout > 0) ? timeout : Long.MAX_VALUE);
while (blockFile != null && blockFile.exists()) {
if (failtime < Time.now()) {
if (failtime < Time.monotonicNow()) {
throw new TimeoutException("waited too long for blocks to be deleted: "
+ blockFile.getPath() + (blockFile.exists() ? " still exists; " : " is absent; "));
}
@ -462,7 +462,7 @@ public class TestDatanodeBlockScanner {
@Test
public void testDuplicateScans() throws Exception {
long startTime = Time.now();
long startTime = Time.monotonicNow();
MiniDFSCluster cluster = new MiniDFSCluster.Builder(new Configuration())
.numDataNodes(1).build();
FileSystem fs = null;