Author: stack
Date: Tue Mar 4 11:25:14 2008
New Revision: 633597
URL: http://svn.apache.org/viewvc?rev=633597&view=rev
Log:
HBASE-490 Doubly-assigned .META.; master uses one and clients another
Modified:
hadoop/hbase/branches/0.1/CHANGES.txt
hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/HMaster.java
hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/util/Sleeper.java
hadoop/hbase/trunk/CHANGES.txt
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/BaseScanner.java
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/MetaScanner.java
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RegionManager.java
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RootScanner.java
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/util/Sleeper.java
Modified: hadoop/hbase/branches/0.1/CHANGES.txt
URL:
http://svn.apache.org/viewvc/hadoop/hbase/branches/0.1/CHANGES.txt?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
--- hadoop/hbase/branches/0.1/CHANGES.txt (original)
+++ hadoop/hbase/branches/0.1/CHANGES.txt Tue Mar 4 11:25:14 2008
@@ -28,6 +28,7 @@
HBASE-446 Fully qualified hbase.rootdir doesn't work
HBASE-428 Under continuous upload of rows, WrongRegionExceptions are
thrown that reach the client even after retries
+ HBASE-490 Doubly-assigned .META.; master uses one and clients another
IMPROVEMENTS
HADOOP-2555 Refactor the HTable#get and HTable#getRow methods to avoid
Modified:
hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/HMaster.java
URL:
http://svn.apache.org/viewvc/hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/HMaster.java?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
--- hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/HMaster.java
(original)
+++ hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/HMaster.java Tue
Mar 4 11:25:14 2008
@@ -407,11 +407,11 @@
}
protected void checkAssigned(final HRegionInfo info,
- final String serverName, final long startCode) throws IOException {
-
+ final String serverName, final long startCode)
+ throws IOException {
// Skip region - if ...
- if(info.isOffline() // offline
- || killedRegions.contains(info.getRegionName()) // queued for offline
+ if (info.isOffline() // offline
+ || killedRegions.contains(info.getRegionName()) // queued for
offline
|| regionsToDelete.contains(info.getRegionName())) { // queued for
delete
unassignedRegions.remove(info);
@@ -424,9 +424,8 @@
Map<Text, HRegionInfo> regionsToKill = killList.get(serverName);
if (regionsToKill != null &&
regionsToKill.containsKey(info.getRegionName())) {
-
// Skip if region is on kill list
- if(LOG.isDebugEnabled()) {
+ if (LOG.isDebugEnabled()) {
LOG.debug("not assigning region (on kill list): " +
info.getRegionName());
}
@@ -438,21 +437,13 @@
}
/*
- * If the server is not dead and either:
- * the stored info is not null and the start code does not match
- * or:
- * the stored info is null and the region is neither unassigned nor
pending
- * then:
+ * If the server is a dead server or its startcode is off -- either null
+ * or doesn't match the start code for the address -- then add it to the
+ * list of unassigned regions IF not already there (or pending open).
*/
- if (!deadServer &&
- ((storedInfo != null && storedInfo.getStartCode() != startCode) ||
- (storedInfo == null &&
- !unassignedRegions.containsKey(info) &&
- !pendingRegions.contains(info.getRegionName())
- )
- )
- ) {
-
+ if (!deadServer && !unassignedRegions.containsKey(info) &&
+ !pendingRegions.contains(info.getRegionName())
+ && (storedInfo == null || storedInfo.getStartCode() != startCode)) {
// The current assignment is invalid
if (LOG.isDebugEnabled()) {
LOG.debug("Current assignment of " + info.getRegionName() +
@@ -503,13 +494,13 @@
super(true, metaRescanInterval, closed);
}
+ // Don't retry if we get an error while scanning. Errors are most often
+ // caused by the server going away. Wait until next rescan interval when
+ // things should be back to normal
private boolean scanRoot() {
- // Don't retry if we get an error while scanning. Errors are most often
- // caused by the server going away. Wait until next rescan interval when
- // things should be back to normal
boolean scanSuccessful = false;
synchronized (rootRegionLocation) {
- while(!closed.get() && rootRegionLocation.get() == null) {
+ while (!closed.get() && rootRegionLocation.get() == null) {
// rootRegionLocation will be filled in when we get an 'open region'
// regionServerReport message from the HRegionServer that has been
// allocated the ROOT region below.
@@ -639,7 +630,7 @@
/** Set by root scanner to indicate the number of meta regions */
volatile AtomicInteger numberOfMetaRegions = new AtomicInteger();
- /** Work for the meta scanner is queued up here */
+ /** Initial work for the meta scanner is queued up here */
volatile BlockingQueue<MetaRegion> metaRegionsToScan =
new LinkedBlockingQueue<MetaRegion>();
@@ -668,10 +659,10 @@
super(false, metaRescanInterval, closed);
}
+ // Don't retry if we get an error while scanning. Errors are most often
+ // caused by the server going away. Wait until next rescan interval when
+ // things should be back to normal
private boolean scanOneMetaRegion(MetaRegion region) {
- // Don't retry if we get an error while scanning. Errors are most often
- // caused by the server going away. Wait until next rescan interval when
- // things should be back to normal
boolean scanSuccessful = false;
while (!closed.get() && !rootScanned &&
rootRegionLocation.get() == null) {
@@ -713,7 +704,12 @@
@Override
protected boolean initialScan() {
MetaRegion region = null;
- while (!closed.get() && region == null && !metaRegionsScanned()) {
+ // Keep going if not closed, metaRegionsToScan has been emptied (or it
+ // hasn't gotten anything in it yet) and all meta regions are onlined
+ // (root and meta).
+ while (!closed.get() &&
+ (region == null || metaRegionsToScan.size() > 0) &&
+ !metaRegionsScanned()) {
try {
region =
metaRegionsToScan.poll(threadWakeFrequency, TimeUnit.MILLISECONDS);
Modified:
hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/util/Sleeper.java
URL:
http://svn.apache.org/viewvc/hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/util/Sleeper.java?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
---
hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/util/Sleeper.java
(original)
+++
hadoop/hbase/branches/0.1/src/java/org/apache/hadoop/hbase/util/Sleeper.java
Tue Mar 4 11:25:14 2008
@@ -44,7 +44,7 @@
* Sleep for period.
*/
public void sleep() {
- sleep(period);
+ sleep(System.currentTimeMillis());
}
/**
Modified: hadoop/hbase/trunk/CHANGES.txt
URL:
http://svn.apache.org/viewvc/hadoop/hbase/trunk/CHANGES.txt?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
--- hadoop/hbase/trunk/CHANGES.txt (original)
+++ hadoop/hbase/trunk/CHANGES.txt Tue Mar 4 11:25:14 2008
@@ -28,6 +28,7 @@
HBASE-462 Update migration tool
HBASE-473 When a table is deleted, master sends multiple close messages to
the region server
+ HBASE-490 Doubly-assigned .META.; master uses one and clients another
IMPROVEMENTS
HBASE-415 Rewrite leases to use DelayedBlockingQueue instead of polling
Modified:
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/BaseScanner.java
URL:
http://svn.apache.org/viewvc/hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/BaseScanner.java?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
--- hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/BaseScanner.java
(original)
+++ hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/BaseScanner.java
Tue Mar 4 11:25:14 2008
@@ -372,21 +372,13 @@
}
/*
- * If the server is not dead and either:
- * the stored info is not null and the start code does not match
- * or:
- * the stored info is null and the region is neither unassigned nor
pending
- * then:
+ * If the server is a dead server or its startcode is off -- either null
+ * or doesn't match the start code for the address -- then add it to the
+ * list of unassigned regions IF not already there (or pending open).
*/
- if (!deadServer &&
- ((storedInfo != null && storedInfo.getStartCode() != startCode) ||
- (storedInfo == null &&
- !regionManager.isUnassigned(info) &&
+ if (!deadServer && !regionManager.isUnassigned(info) &&
!regionManager.isPending(info.getRegionName())
- )
- )
- ) {
-
+ && (storedInfo == null || storedInfo.getStartCode() != startCode)) {
// The current assignment is invalid
if (LOG.isDebugEnabled()) {
LOG.debug("Current assignment of " + info.getRegionName() +
Modified:
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/MetaScanner.java
URL:
http://svn.apache.org/viewvc/hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/MetaScanner.java?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
--- hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/MetaScanner.java
(original)
+++ hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/MetaScanner.java
Tue Mar 4 11:25:14 2008
@@ -38,7 +38,7 @@
* action would prevent other work from getting done.
*/
class MetaScanner extends BaseScanner {
- /** Work for the meta scanner is queued up here */
+ /** Initial work for the meta scanner is queued up here */
private volatile BlockingQueue<MetaRegion> metaRegionsToScan =
new LinkedBlockingQueue<MetaRegion>();
@@ -50,10 +50,10 @@
super(master, regionManager, false, master.metaRescanInterval,
master.closed);
}
+ // Don't retry if we get an error while scanning. Errors are most often
+ // caused by the server going away. Wait until next rescan interval when
+ // things should be back to normal
private boolean scanOneMetaRegion(MetaRegion region) {
- // Don't retry if we get an error while scanning. Errors are most often
- // caused by the server going away. Wait until next rescan interval when
- // things should be back to normal
boolean scanSuccessful = false;
while (!master.closed.get() && !regionManager.isInitialRootScanComplete()
&&
regionManager.getRootRegionLocation() == null) {
@@ -95,7 +95,9 @@
@Override
protected boolean initialScan() {
MetaRegion region = null;
- while (!master.closed.get() && region == null && !metaRegionsScanned()) {
+ while (!master.closed.get() &&
+ (region == null && metaRegionsToScan.size() > 0) &&
+ !metaRegionsScanned()) {
try {
region = metaRegionsToScan.poll(master.threadWakeFrequency,
TimeUnit.MILLISECONDS);
@@ -164,4 +166,4 @@
void addMetaRegionToScan(MetaRegion m) throws InterruptedException {
metaRegionsToScan.add(m);
}
-}
\ No newline at end of file
+}
Modified:
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RegionManager.java
URL:
http://svn.apache.org/viewvc/hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RegionManager.java?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
---
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RegionManager.java
(original)
+++
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RegionManager.java
Tue Mar 4 11:25:14 2008
@@ -598,7 +598,7 @@
public void waitForRootRegionLocation() {
synchronized (rootRegionLocation) {
- while(!master.closed.get() && rootRegionLocation.get() == null) {
+ while (!master.closed.get() && rootRegionLocation.get() == null) {
// rootRegionLocation will be filled in when we get an 'open region'
// regionServerReport message from the HRegionServer that has been
// allocated the ROOT region below.
Modified:
hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RootScanner.java
URL:
http://svn.apache.org/viewvc/hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RootScanner.java?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
--- hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RootScanner.java
(original)
+++ hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/master/RootScanner.java
Tue Mar 4 11:25:14 2008
@@ -31,10 +31,10 @@
super(master, regionManager, true, master.metaRescanInterval,
master.closed);
}
+ // Don't retry if we get an error while scanning. Errors are most often
+ // caused by the server going away. Wait until next rescan interval when
+ // things should be back to normal
private boolean scanRoot() {
- // Don't retry if we get an error while scanning. Errors are most often
- // caused by the server going away. Wait until next rescan interval when
- // things should be back to normal
boolean scanSuccessful = false;
master.waitForRootRegionLocation();
if (master.closed.get()) {
@@ -71,4 +71,4 @@
protected void maintenanceScan() {
scanRoot();
}
-}
\ No newline at end of file
+}
Modified: hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/util/Sleeper.java
URL:
http://svn.apache.org/viewvc/hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/util/Sleeper.java?rev=633597&r1=633596&r2=633597&view=diff
==============================================================================
--- hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/util/Sleeper.java
(original)
+++ hadoop/hbase/trunk/src/java/org/apache/hadoop/hbase/util/Sleeper.java Tue
Mar 4 11:25:14 2008
@@ -44,7 +44,7 @@
* Sleep for period.
*/
public void sleep() {
- sleep(period);
+ sleep(System.currentTimeMillis());
}
/**
@@ -69,4 +69,4 @@
}
}
}
-}
\ No newline at end of file
+}
