svn merge -c 1469839 Merging from trunk to branch-2 to fix HDFS-4699.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/branch-2@1469840 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Kihwal Lee 2013-04-19 14:11:28 +00:00
parent 1bfb6af49f
commit fbce046a88
4 changed files with 22 additions and 6 deletions

View File

@ -50,6 +50,9 @@ Release 2.0.5-beta - UNRELEASED
HDFS-4693. Some test cases in TestCheckpoint do not clean up after
themselves. (Arpit Agarwal, suresh via suresh)
HDFS-3817. Avoid printing SafeModeException stack trace.
(Brandon Li via suresh)
OPTIMIZATIONS
BUG FIXES
@ -2221,6 +2224,9 @@ Release 0.23.8 - UNRELEASED
HDFS-4477. Secondary namenode may retain old tokens (daryn via kihwal)
HDFS-4699. TestPipelinesFailover#testPipelineRecoveryStress fails
sporadically (Chris Nauroth via kihwal)
Release 0.23.7 - UNRELEASED
INCOMPATIBLE CHANGES

View File

@ -1177,7 +1177,10 @@ public class DataNode extends Configured
LOG.warn("checkDiskError: exception: ", e);
if (e instanceof SocketException || e instanceof SocketTimeoutException
|| e instanceof ClosedByInterruptException
|| e.getMessage().startsWith("Broken pipe")) {
|| e.getMessage().startsWith("An established connection was aborted")
|| e.getMessage().startsWith("Broken pipe")
|| e.getMessage().startsWith("Connection reset")
|| e.getMessage().contains("java.nio.channels.SocketChannel")) {
LOG.info("Not checking disk as checkDiskError was called on a network" +
" related exception");
return;

View File

@ -273,7 +273,10 @@ class NameNodeRpcServer implements NamenodeProtocols {
minimumDataNodeVersion = conf.get(
DFSConfigKeys.DFS_NAMENODE_MIN_SUPPORTED_DATANODE_VERSION_KEY,
DFSConfigKeys.DFS_NAMENODE_MIN_SUPPORTED_DATANODE_VERSION_DEFAULT);
}
// Set terse exception whose stack trace won't be logged
this.clientRpcServer.addTerseExceptions(SafeModeException.class);
}
/**
* Start client and service RPC servers.

View File

@ -422,6 +422,11 @@ public class TestPipelinesFailover {
// Disable permissions so that another user can recover the lease.
harness.conf.setBoolean(
DFSConfigKeys.DFS_PERMISSIONS_ENABLED_KEY, false);
// This test triggers rapid NN failovers. The client retry policy uses an
// exponential backoff. This can quickly lead to long sleep times and even
// timeout the whole test. Cap the sleep time at 1s to prevent this.
harness.conf.setInt(DFSConfigKeys.DFS_CLIENT_FAILOVER_SLEEPTIME_MAX_KEY,
1000);
final MiniDFSCluster cluster = harness.startCluster();
try {
@ -537,11 +542,10 @@ public class TestPipelinesFailover {
}
/**
* Try to cover the lease on the given file for up to 30
* seconds.
* Try to recover the lease on the given file for up to 60 seconds.
* @param fsOtherUser the filesystem to use for the recoverLease call
* @param testPath the path on which to run lease recovery
* @throws TimeoutException if lease recover does not succeed within 30
* @throws TimeoutException if lease recover does not succeed within 60
* seconds
* @throws InterruptedException if the thread is interrupted
*/
@ -564,7 +568,7 @@ public class TestPipelinesFailover {
}
return success;
}
}, 1000, 30000);
}, 1000, 60000);
} catch (TimeoutException e) {
throw new TimeoutException("Timed out recovering lease for " +
testPath);