HDFS-2577. NN fails to start since it tries to start secret manager in safemode. Contributed by Todd Lipcon.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/HDFS-1623@1205689 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Todd Lipcon 2011-11-24 00:55:37 +00:00
parent 6a0671977b
commit 5f5b3bbdcd
5 changed files with 65 additions and 23 deletions

View File

@ -21,3 +21,5 @@ HDFS-2418. Change ConfiguredFailoverProxyProvider to take advantage of HDFS-2231
HDFS-2393. Mark appropriate methods of ClientProtocol with the idempotent annotation. (atm)
HDFS-2523. Small NN fixes to include HAServiceProtocol and prevent NPE on shutdown. (todd)
HDFS-2577. NN fails to start since it tries to start secret manager in safemode. (todd)

View File

@ -258,7 +258,6 @@ private static final void logAuditEvent(UserGroupInformation ugi,
LeaseManager leaseManager = new LeaseManager(this);
Daemon lmthread = null; // LeaseMonitor thread
Daemon smmthread = null; // SafeModeMonitor thread
Daemon nnrmthread = null; // NamenodeResourceMonitor thread
@ -450,9 +449,10 @@ void startActiveServices() throws IOException {
LOG.info("Starting services required for active state");
writeLock();
try {
startSecretManager();
lmthread = new Daemon(leaseManager.new Monitor());
lmthread.start();
if (UserGroupInformation.isSecurityEnabled()) {
startSecretManager();
}
leaseManager.startMonitor();
} finally {
writeUnlock();
}
@ -467,14 +467,8 @@ void stopActiveServices() {
writeLock();
try {
stopSecretManager();
if (lmthread != null) {
try {
lmthread.interrupt();
lmthread.join(3000);
} catch (InterruptedException ie) {
LOG.warn("Encountered exception ", ie);
}
lmthread = null;
if (leaseManager != null) {
leaseManager.stopMonitor();
}
} finally {
writeUnlock();
@ -542,6 +536,10 @@ public void writeLock() {
this.fsLock.writeLock().lock();
}
@Override
public void writeLockInterruptibly() throws InterruptedException {
this.fsLock.writeLock().lockInterruptibly();
}
@Override
public void writeUnlock() {
this.fsLock.writeLock().unlock();
}

View File

@ -34,6 +34,10 @@
import org.apache.hadoop.fs.UnresolvedLinkException;
import org.apache.hadoop.hdfs.protocol.HdfsConstants;
import org.apache.hadoop.hdfs.server.common.HdfsServerConstants;
import org.apache.hadoop.util.Daemon;
import com.google.common.annotations.VisibleForTesting;
import com.google.common.base.Preconditions;
import static org.apache.hadoop.hdfs.server.common.Util.now;
@ -82,6 +86,9 @@ public class LeaseManager {
//
private SortedMap<String, Lease> sortedLeasesByPath = new TreeMap<String, Lease>();
private Daemon lmthread;
private volatile boolean shouldRunMonitor;
LeaseManager(FSNamesystem fsnamesystem) {this.fsnamesystem = fsnamesystem;}
Lease getLease(String holder) {
@ -367,18 +374,18 @@ class Monitor implements Runnable {
/** Check leases periodically. */
public void run() {
for(; fsnamesystem.isRunning(); ) {
fsnamesystem.writeLock();
for(; shouldRunMonitor && fsnamesystem.isRunning(); ) {
try {
if (!fsnamesystem.isInSafeMode()) {
checkLeases();
fsnamesystem.writeLockInterruptibly();
try {
if (!fsnamesystem.isInSafeMode()) {
checkLeases();
}
} finally {
fsnamesystem.writeUnlock();
}
} finally {
fsnamesystem.writeUnlock();
}
try {
Thread.sleep(HdfsServerConstants.NAMENODE_LEASE_RECHECK_INTERVAL);
} catch(InterruptedException ie) {
if (LOG.isDebugEnabled()) {
@ -437,4 +444,36 @@ public synchronized String toString() {
+ "\n sortedLeasesByPath=" + sortedLeasesByPath
+ "\n}";
}
void startMonitor() {
Preconditions.checkState(lmthread == null,
"Lease Monitor already running");
shouldRunMonitor = true;
lmthread = new Daemon(new Monitor());
lmthread.start();
}
void stopMonitor() {
if (lmthread != null) {
shouldRunMonitor = false;
try {
lmthread.interrupt();
lmthread.join(3000);
} catch (InterruptedException ie) {
LOG.warn("Encountered exception ", ie);
}
lmthread = null;
}
}
/**
* Trigger the currently-running Lease monitor to re-check
* its leases immediately. This is for use by unit tests.
*/
@VisibleForTesting
void triggerMonitorCheckNow() {
Preconditions.checkState(lmthread != null,
"Lease monitor is not running");
lmthread.interrupt();
}
}

View File

@ -30,6 +30,9 @@ public interface RwLock {
/** Acquire write lock. */
public void writeLock();
/** Acquire write lock, unless interrupted while waiting */
void writeLockInterruptibly() throws InterruptedException;
/** Release write lock. */
public void writeUnlock();

View File

@ -78,7 +78,7 @@ public static LeaseManager getLeaseManager(final FSNamesystem ns) {
/** Set the softLimit and hardLimit of client lease periods. */
public static void setLeasePeriod(final FSNamesystem namesystem, long soft, long hard) {
getLeaseManager(namesystem).setLeasePeriod(soft, hard);
namesystem.lmthread.interrupt();
namesystem.leaseManager.triggerMonitorCheckNow();
}
public static String getLeaseHolderForPath(NameNode namenode, String path) {