HDFS-5159. Secondary NameNode fails to checkpoint if error occurs downloading edits on first checkpoint. Contributed by Aaron T. Myers.
git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1520363 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
parent
e8f6f74025
commit
8aea748ec3
@ -415,6 +415,9 @@ Release 2.1.1-beta - UNRELEASED
|
||||
HDFS-5140. Too many safemode monitor threads being created in the standby
|
||||
namenode causing it to fail with out of memory error. (jing9)
|
||||
|
||||
HDFS-5159. Secondary NameNode fails to checkpoint if error occurs
|
||||
downloading edits on first checkpoint. (atm)
|
||||
|
||||
Release 2.1.0-beta - 2013-08-22
|
||||
|
||||
INCOMPATIBLE CHANGES
|
||||
|
@ -429,10 +429,8 @@ public Boolean run() throws Exception {
|
||||
dstImage.getStorage().cTime = sig.cTime;
|
||||
|
||||
// get fsimage
|
||||
boolean downloadImage = true;
|
||||
if (sig.mostRecentCheckpointTxId ==
|
||||
dstImage.getStorage().getMostRecentCheckpointTxId()) {
|
||||
downloadImage = false;
|
||||
LOG.info("Image has not changed. Will not download image.");
|
||||
} else {
|
||||
LOG.info("Image has changed. Downloading updated image from NN.");
|
||||
@ -448,7 +446,9 @@ public Boolean run() throws Exception {
|
||||
nnHostPort, log, dstImage.getStorage());
|
||||
}
|
||||
|
||||
return Boolean.valueOf(downloadImage);
|
||||
// true if we haven't loaded all the transactions represented by the
|
||||
// downloaded fsimage.
|
||||
return dstImage.getLastAppliedTxId() < sig.mostRecentCheckpointTxId;
|
||||
}
|
||||
});
|
||||
return b.booleanValue();
|
||||
|
@ -39,7 +39,6 @@
|
||||
import java.util.List;
|
||||
|
||||
import org.apache.commons.cli.ParseException;
|
||||
import org.apache.commons.io.filefilter.FileFilterUtils;
|
||||
import org.apache.commons.logging.Log;
|
||||
import org.apache.commons.logging.LogFactory;
|
||||
import org.apache.commons.logging.impl.Log4JLogger;
|
||||
@ -1224,7 +1223,6 @@ public void testSaveNamespace() throws IOException {
|
||||
}
|
||||
|
||||
/* Test case to test CheckpointSignature */
|
||||
@SuppressWarnings("deprecation")
|
||||
@Test
|
||||
public void testCheckpointSignature() throws IOException {
|
||||
|
||||
@ -1562,12 +1560,65 @@ public void testEditFailureBeforeRename() throws IOException {
|
||||
Mockito.reset(faultInjector);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Test that a fault while downloading edits the first time after the 2NN
|
||||
* starts up does not prevent future checkpointing.
|
||||
*/
|
||||
@Test(timeout = 30000)
|
||||
public void testEditFailureOnFirstCheckpoint() throws IOException {
|
||||
Configuration conf = new HdfsConfiguration();
|
||||
SecondaryNameNode secondary = null;
|
||||
MiniDFSCluster cluster = null;
|
||||
FileSystem fs = null;
|
||||
try {
|
||||
cluster = new MiniDFSCluster.Builder(conf).numDataNodes(numDatanodes)
|
||||
.build();
|
||||
cluster.waitActive();
|
||||
fs = cluster.getFileSystem();
|
||||
fs.mkdirs(new Path("test-file-1"));
|
||||
|
||||
// Make sure the on-disk fsimage on the NN has txid > 0.
|
||||
FSNamesystem fsns = cluster.getNamesystem();
|
||||
fsns.enterSafeMode(false);
|
||||
fsns.saveNamespace();
|
||||
fsns.leaveSafeMode();
|
||||
|
||||
secondary = startSecondaryNameNode(conf);
|
||||
|
||||
// Cause edit rename to fail during next checkpoint
|
||||
Mockito.doThrow(new IOException("Injecting failure before edit rename"))
|
||||
.when(faultInjector).beforeEditsRename();
|
||||
|
||||
try {
|
||||
secondary.doCheckpoint();
|
||||
fail("Fault injection failed.");
|
||||
} catch (IOException ioe) {
|
||||
GenericTestUtils.assertExceptionContains(
|
||||
"Injecting failure before edit rename", ioe);
|
||||
}
|
||||
Mockito.reset(faultInjector);
|
||||
|
||||
// Next checkpoint should succeed
|
||||
secondary.doCheckpoint();
|
||||
} finally {
|
||||
if (secondary != null) {
|
||||
secondary.shutdown();
|
||||
}
|
||||
if (fs != null) {
|
||||
fs.close();
|
||||
}
|
||||
if (cluster != null) {
|
||||
cluster.shutdown();
|
||||
}
|
||||
Mockito.reset(faultInjector);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Test that the secondary namenode correctly deletes temporary edits
|
||||
* on startup.
|
||||
*/
|
||||
|
||||
@Test(timeout = 30000)
|
||||
public void testDeleteTemporaryEditsOnStartup() throws IOException {
|
||||
Configuration conf = new HdfsConfiguration();
|
||||
@ -1943,7 +1994,6 @@ public void testNamespaceVerifiedOnFileTransfer() throws IOException {
|
||||
* Test that, if a storage directory is failed when a checkpoint occurs,
|
||||
* the non-failed storage directory receives the checkpoint.
|
||||
*/
|
||||
@SuppressWarnings("deprecation")
|
||||
@Test
|
||||
public void testCheckpointWithFailedStorageDir() throws Exception {
|
||||
MiniDFSCluster cluster = null;
|
||||
@ -2006,7 +2056,6 @@ public void testCheckpointWithFailedStorageDir() throws Exception {
|
||||
* should function correctly.
|
||||
* @throws Exception
|
||||
*/
|
||||
@SuppressWarnings("deprecation")
|
||||
@Test
|
||||
public void testCheckpointWithSeparateDirsAfterNameFails() throws Exception {
|
||||
MiniDFSCluster cluster = null;
|
||||
|
Loading…
Reference in New Issue
Block a user