Author: umamahesh
Date: Tue May 29 19:05:40 2012
New Revision: 1343929
URL: http://svn.apache.org/viewvc?rev=1343929&view=rev
Log:
Merge r:1343913 HDFS-3452. BKJM:Switch from standby to active fails and NN gets
shut down due to delay in clearing of lock. Contributed by Uma Maheswara Rao G.
Added:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/CurrentInprogress.java
- copied unchanged from r1343913,
hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/CurrentInprogress.java
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestCurrentInprogress.java
- copied unchanged from r1343913,
hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestCurrentInprogress.java
Removed:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/WriteLock.java
Modified:
hadoop/common/branches/branch-2/hadoop-hdfs-project/ (props changed)
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/ (props
changed)
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperEditLogOutputStream.java
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperJournalManager.java
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperAsHASharedDir.java
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperJournalManager.java
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/java/
(props changed)
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/native/
(props changed)
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/datanode/
(props changed)
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/hdfs/
(props changed)
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/secondary/
(props changed)
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/test/hdfs/
(props changed)
Propchange: hadoop/common/branches/branch-2/hadoop-hdfs-project/
------------------------------------------------------------------------------
Merged /hadoop/common/trunk/hadoop-hdfs-project:r1343913
Propchange: hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/
------------------------------------------------------------------------------
Merged /hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs:r1343913
Modified:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
URL:
http://svn.apache.org/viewvc/hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt?rev=1343929&r1=1343928&r2=1343929&view=diff
==============================================================================
--- hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
(original)
+++ hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
Tue May 29 19:05:40 2012
@@ -118,6 +118,9 @@ Release 2.0.1-alpha - UNRELEASED
HDFS-3368. Missing blocks due to bad DataNodes coming up and down. (shv)
+ HDFS-3452. BKJM:Switch from standby to active fails and NN gets shut down
+ due to delay in clearing of lock. (umamahesh)
+
Release 2.0.0-alpha - UNRELEASED
INCOMPATIBLE CHANGES
Modified:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperEditLogOutputStream.java
URL:
http://svn.apache.org/viewvc/hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperEditLogOutputStream.java?rev=1343929&r1=1343928&r2=1343929&view=diff
==============================================================================
---
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperEditLogOutputStream.java
(original)
+++
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperEditLogOutputStream.java
Tue May 29 19:05:40 2012
@@ -56,15 +56,13 @@ class BookKeeperEditLogOutputStream
private CountDownLatch syncLatch;
private final AtomicInteger transmitResult
= new AtomicInteger(BKException.Code.OK);
- private final WriteLock wl;
private final Writer writer;
/**
* Construct an edit log output stream which writes to a ledger.
*/
- protected BookKeeperEditLogOutputStream(Configuration conf,
- LedgerHandle lh, WriteLock wl)
+ protected BookKeeperEditLogOutputStream(Configuration conf, LedgerHandle lh)
throws IOException {
super();
@@ -72,8 +70,6 @@ class BookKeeperEditLogOutputStream
outstandingRequests = new AtomicInteger(0);
syncLatch = null;
this.lh = lh;
- this.wl = wl;
- this.wl.acquire();
this.writer = new Writer(bufCurrent);
this.transmissionThreshold
= conf.getInt(BookKeeperJournalManager.BKJM_OUTPUT_BUFFER_SIZE,
@@ -108,7 +104,6 @@ class BookKeeperEditLogOutputStream
throw new IOException("BookKeeper error during abort", bke);
}
- wl.release();
}
@Override
@@ -118,8 +113,6 @@ class BookKeeperEditLogOutputStream
@Override
public void write(FSEditLogOp op) throws IOException {
- wl.checkWriteLock();
-
writer.writeOp(op);
if (bufCurrent.getLength() > transmissionThreshold) {
@@ -129,19 +122,15 @@ class BookKeeperEditLogOutputStream
@Override
public void setReadyToFlush() throws IOException {
- wl.checkWriteLock();
-
transmit();
- synchronized(this) {
+ synchronized (this) {
syncLatch = new CountDownLatch(outstandingRequests.get());
}
}
@Override
public void flushAndSync() throws IOException {
- wl.checkWriteLock();
-
assert(syncLatch != null);
try {
syncLatch.await();
@@ -164,8 +153,6 @@ class BookKeeperEditLogOutputStream
* are never called at the same time.
*/
private void transmit() throws IOException {
- wl.checkWriteLock();
-
if (!transmitResult.compareAndSet(BKException.Code.OK,
BKException.Code.OK)) {
throw new IOException("Trying to write to an errored stream;"
Modified:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperJournalManager.java
URL:
http://svn.apache.org/viewvc/hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperJournalManager.java?rev=1343929&r1=1343928&r2=1343929&view=diff
==============================================================================
---
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperJournalManager.java
(original)
+++
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/main/java/org/apache/hadoop/contrib/bkjournal/BookKeeperJournalManager.java
Tue May 29 19:05:40 2012
@@ -117,7 +117,7 @@ public class BookKeeperJournalManager im
private final ZooKeeper zkc;
private final Configuration conf;
private final BookKeeper bkc;
- private final WriteLock wl;
+ private final CurrentInprogress ci;
private final String ledgerPath;
private final MaxTxId maxTxId;
private final int ensembleSize;
@@ -155,7 +155,7 @@ public class BookKeeperJournalManager im
ledgerPath = zkPath + "/ledgers";
String maxTxIdPath = zkPath + "/maxtxid";
- String lockPath = zkPath + "/lock";
+ String currentInprogressNodePath = zkPath + "/CurrentInprogress";
String versionPath = zkPath + "/version";
digestpw = conf.get(BKJM_BOOKKEEPER_DIGEST_PW,
BKJM_BOOKKEEPER_DIGEST_PW_DEFAULT);
@@ -192,7 +192,7 @@ public class BookKeeperJournalManager im
throw new IOException("Error initializing zk", e);
}
- wl = new WriteLock(zkc, lockPath);
+ ci = new CurrentInprogress(zkc, currentInprogressNodePath);
maxTxId = new MaxTxId(zkc, maxTxIdPath);
}
@@ -207,13 +207,16 @@ public class BookKeeperJournalManager im
*/
@Override
public EditLogOutputStream startLogSegment(long txId) throws IOException {
- wl.acquire();
-
if (txId <= maxTxId.get()) {
throw new IOException("We've already seen " + txId
+ ". A new stream cannot be created with it");
}
try {
+ String existingInprogressNode = ci.read();
+ if (null != existingInprogressNode
+ && zkc.exists(existingInprogressNode, false) != null) {
+ throw new IOException("Inprogress node already exists");
+ }
if (currentLedger != null) {
// bookkeeper errored on last stream, clean up ledger
currentLedger.close();
@@ -234,7 +237,8 @@ public class BookKeeperJournalManager im
l.write(zkc, znodePath);
maxTxId.store(txId);
- return new BookKeeperEditLogOutputStream(conf, currentLedger, wl);
+ ci.update(znodePath);
+ return new BookKeeperEditLogOutputStream(conf, currentLedger);
} catch (Exception e) {
if (currentLedger != null) {
try {
@@ -270,7 +274,6 @@ public class BookKeeperJournalManager im
+ " doesn't exist");
}
- wl.checkWriteLock();
EditLogLedgerMetadata l
= EditLogLedgerMetadata.read(zkc, inprogressPath);
@@ -307,13 +310,15 @@ public class BookKeeperJournalManager im
}
maxTxId.store(lastTxId);
zkc.delete(inprogressPath, inprogressStat.getVersion());
+ String inprogressPathFromCI = ci.read();
+ if (inprogressPath.equals(inprogressPathFromCI)) {
+ ci.clear();
+ }
} catch (KeeperException e) {
throw new IOException("Error finalising ledger", e);
} catch (InterruptedException ie) {
throw new IOException("Error finalising ledger", ie);
- } finally {
- wl.release();
- }
+ }
}
EditLogInputStream getInputStream(long fromTxId, boolean inProgressOk)
@@ -417,7 +422,6 @@ public class BookKeeperJournalManager im
@Override
public void recoverUnfinalizedSegments() throws IOException {
- wl.acquire();
synchronized (this) {
try {
List<String> children = zkc.getChildren(ledgerPath, false);
@@ -445,10 +449,6 @@ public class BookKeeperJournalManager im
} catch (InterruptedException ie) {
throw new IOException("Interrupted getting list of inprogress
segments",
ie);
- } finally {
- if (wl.haveLock()) {
- wl.release();
- }
}
}
}
Modified:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperAsHASharedDir.java
URL:
http://svn.apache.org/viewvc/hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperAsHASharedDir.java?rev=1343929&r1=1343928&r2=1343929&view=diff
==============================================================================
---
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperAsHASharedDir.java
(original)
+++
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperAsHASharedDir.java
Tue May 29 19:05:40 2012
@@ -215,8 +215,7 @@ public class TestBookKeeperAsHASharedDir
}
/**
- * Test that two namenodes can't become primary at the same
- * time.
+ * Test that two namenodes can't continue as primary
*/
@Test
public void testMultiplePrimariesStarted() throws Exception {
@@ -247,21 +246,17 @@ public class TestBookKeeperAsHASharedDir
FileSystem fs = HATestUtil.configureFailoverFs(cluster, conf);
fs.mkdirs(p1);
nn1.getRpcServer().rollEditLog();
- try {
- cluster.transitionToActive(1);
- fail("Shouldn't have been able to start two primaries"
- + " with single shared storage");
- } catch (ServiceFailedException sfe) {
- assertTrue("Wrong exception",
- sfe.getMessage().contains("Failed to start active services"));
- }
+ cluster.transitionToActive(1);
+ fs = cluster.getFileSystem(0); // get the older active server.
+ // This edit log updation on older active should make older active
+ // shutdown.
+ fs.delete(p1, true);
+ verify(mockRuntime1, atLeastOnce()).exit(anyInt());
+ verify(mockRuntime2, times(0)).exit(anyInt());
} finally {
- verify(mockRuntime1, times(0)).exit(anyInt());
- verify(mockRuntime2, atLeastOnce()).exit(anyInt());
-
if (cluster != null) {
cluster.shutdown();
}
}
}
-}
\ No newline at end of file
+}
Modified:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperJournalManager.java
URL:
http://svn.apache.org/viewvc/hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperJournalManager.java?rev=1343929&r1=1343928&r2=1343929&view=diff
==============================================================================
---
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperJournalManager.java
(original)
+++
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/contrib/bkjournal/src/test/java/org/apache/hadoop/contrib/bkjournal/TestBookKeeperJournalManager.java
Tue May 29 19:05:40 2012
@@ -361,6 +361,7 @@ public class TestBookKeeperJournalManage
assertEquals("New bookie didn't start",
numBookies+1, bkutil.checkBookiesUp(numBookies+1, 10));
+ bkjm.recoverUnfinalizedSegments();
out = bkjm.startLogSegment(txid);
for (long i = 1 ; i <= 3; i++) {
FSEditLogOp op = FSEditLogTestUtil.getNoOpInstance();
Propchange:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/java/
------------------------------------------------------------------------------
Merged
/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/java:r1343913
Propchange:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/native/
------------------------------------------------------------------------------
Merged
/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/native:r1343913
Propchange:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/datanode/
------------------------------------------------------------------------------
Merged
/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/datanode:r1343913
Propchange:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/hdfs/
------------------------------------------------------------------------------
Merged
/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/hdfs:r1343913
Propchange:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/secondary/
------------------------------------------------------------------------------
Merged
/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/secondary:r1343913
Propchange:
hadoop/common/branches/branch-2/hadoop-hdfs-project/hadoop-hdfs/src/test/hdfs/
------------------------------------------------------------------------------
Merged
/hadoop/common/trunk/hadoop-hdfs-project/hadoop-hdfs/src/test/hdfs:r1343913
