HBASE-9703 DistributedHBaseCluster should not throw exceptions, but do a best effort restore

git-svn-id: https://svn.apache.org/repos/asf/hbase/trunk@1529045 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Enis Soztutar 2013-10-04 00:55:42 +00:00
parent a07ef888e1
commit 4bde0d30f6
2 changed files with 112 additions and 35 deletions

View File

@ -18,9 +18,9 @@
package org.apache.hadoop.hbase;
import java.io.IOException;
import java.util.ArrayList;
import java.util.HashMap;
import com.google.common.collect.Sets;
import java.util.List;
import org.apache.hadoop.classification.InterfaceAudience;
import org.apache.hadoop.conf.Configuration;
@ -36,6 +36,8 @@ import org.apache.hadoop.hbase.protobuf.generated.MasterProtos.MasterService;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.hbase.util.Threads;
import com.google.common.collect.Sets;
/**
* Manages the interactions with an already deployed distributed cluster (as opposed to
* a pseudo-distributed, or mini/local cluster). This is used by integration and system tests.
@ -215,38 +217,64 @@ public class DistributedHBaseCluster extends HBaseCluster {
}
@Override
public void restoreClusterStatus(ClusterStatus initial) throws IOException {
//TODO: caution: not tested throughly
public boolean restoreClusterStatus(ClusterStatus initial) throws IOException {
ClusterStatus current = getClusterStatus();
//restore masters
LOG.info("Restoring cluster - started");
// do a best effort restore
boolean success = true;
success = restoreMasters(initial, current) & success;
success = restoreRegionServers(initial, current) & success;
success = restoreAdmin() & success;
LOG.info("Restoring cluster - done");
return success;
}
protected boolean restoreMasters(ClusterStatus initial, ClusterStatus current) {
List<IOException> deferred = new ArrayList<IOException>();
//check whether current master has changed
if (!ServerName.isSameHostnameAndPort(initial.getMaster(), current.getMaster())) {
LOG.info("Initial active master : " + initial.getMaster().getHostname()
LOG.info("Restoring cluster - Initial active master : " + initial.getMaster().getHostname()
+ " has changed to : " + current.getMaster().getHostname());
// If initial master is stopped, start it, before restoring the state.
// It will come up as a backup master, if there is already an active master.
if (!clusterManager.isRunning(ServiceType.HBASE_MASTER, initial.getMaster().getHostname())) {
startMaster(initial.getMaster().getHostname());
try {
if (!clusterManager.isRunning(ServiceType.HBASE_MASTER, initial.getMaster().getHostname())) {
LOG.info("Restoring cluster - starting initial active master at:" + initial.getMaster().getHostname());
startMaster(initial.getMaster().getHostname());
}
//master has changed, we would like to undo this.
//1. Kill the current backups
//2. Stop current master
//3. Start backup masters
for (ServerName currentBackup : current.getBackupMasters()) {
if (!ServerName.isSameHostnameAndPort(currentBackup, initial.getMaster())) {
LOG.info("Restoring cluster - stopping backup master: " + currentBackup);
stopMaster(currentBackup);
}
}
LOG.info("Restoring cluster - stopping active master: " + current.getMaster());
stopMaster(current.getMaster());
waitForActiveAndReadyMaster(); //wait so that active master takes over
} catch (IOException ex) {
// if we fail to start the initial active master, we do not want to continue stopping
// backup masters. Just keep what we have now
deferred.add(ex);
}
//master has changed, we would like to undo this.
//1. Kill the current backups
//2. Stop current master
//3. Start backup masters
for (ServerName currentBackup : current.getBackupMasters()) {
if (!ServerName.isSameHostnameAndPort(currentBackup, initial.getMaster())) {
stopMaster(currentBackup);
}
}
stopMaster(current.getMaster());
waitForActiveAndReadyMaster(); //wait so that active master takes over
//start backup masters
for (ServerName backup : initial.getBackupMasters()) {
//these are not started in backup mode, but we should already have an active master
if(!clusterManager.isRunning(ServiceType.HBASE_MASTER, backup.getHostname())) {
startMaster(backup.getHostname());
try {
//these are not started in backup mode, but we should already have an active master
if(!clusterManager.isRunning(ServiceType.HBASE_MASTER, backup.getHostname())) {
LOG.info("Restoring cluster - starting initial backup master: " + backup.getHostname());
startMaster(backup.getHostname());
}
} catch (IOException ex) {
deferred.add(ex);
}
}
} else {
@ -262,19 +290,38 @@ public class DistributedHBaseCluster extends HBaseCluster {
}
for (String hostname : Sets.difference(initialBackups.keySet(), currentBackups.keySet())) {
if(!clusterManager.isRunning(ServiceType.HBASE_MASTER, hostname)) {
startMaster(hostname);
try {
if(!clusterManager.isRunning(ServiceType.HBASE_MASTER, hostname)) {
LOG.info("Restoring cluster - starting initial backup master: " + hostname);
startMaster(hostname);
}
} catch (IOException ex) {
deferred.add(ex);
}
}
for (String hostname : Sets.difference(currentBackups.keySet(), initialBackups.keySet())) {
if(clusterManager.isRunning(ServiceType.HBASE_MASTER, hostname)) {
stopMaster(currentBackups.get(hostname));
try {
if(clusterManager.isRunning(ServiceType.HBASE_MASTER, hostname)) {
LOG.info("Restoring cluster - stopping backup master: " + hostname);
stopMaster(currentBackups.get(hostname));
}
} catch (IOException ex) {
deferred.add(ex);
}
}
}
if (!deferred.isEmpty()) {
LOG.warn("Restoring cluster - restoring region servers reported " + deferred.size() + " errors:");
for (int i=0; i<deferred.size() && i < 3; i++) {
LOG.warn(deferred.get(i));
}
}
//restore region servers
return deferred.isEmpty();
}
protected boolean restoreRegionServers(ClusterStatus initial, ClusterStatus current) {
HashMap<String, ServerName> initialServers = new HashMap<String, ServerName>();
HashMap<String, ServerName> currentServers = new HashMap<String, ServerName>();
@ -285,17 +332,39 @@ public class DistributedHBaseCluster extends HBaseCluster {
currentServers.put(server.getHostname(), server);
}
List<IOException> deferred = new ArrayList<IOException>();
for (String hostname : Sets.difference(initialServers.keySet(), currentServers.keySet())) {
if(!clusterManager.isRunning(ServiceType.HBASE_REGIONSERVER, hostname)) {
startRegionServer(hostname);
try {
if(!clusterManager.isRunning(ServiceType.HBASE_REGIONSERVER, hostname)) {
LOG.info("Restoring cluster - starting initial region server: " + hostname);
startRegionServer(hostname);
}
} catch (IOException ex) {
deferred.add(ex);
}
}
for (String hostname : Sets.difference(currentServers.keySet(), initialServers.keySet())) {
if(clusterManager.isRunning(ServiceType.HBASE_REGIONSERVER, hostname)) {
stopRegionServer(currentServers.get(hostname));
try {
if(clusterManager.isRunning(ServiceType.HBASE_REGIONSERVER, hostname)) {
LOG.info("Restoring cluster - stopping initial region server: " + hostname);
stopRegionServer(currentServers.get(hostname));
}
} catch (IOException ex) {
deferred.add(ex);
}
}
if (!deferred.isEmpty()) {
LOG.warn("Restoring cluster - restoring region servers reported " + deferred.size() + " errors:");
for (int i=0; i<deferred.size() && i < 3; i++) {
LOG.warn(deferred.get(i));
}
}
return deferred.isEmpty();
}
protected boolean restoreAdmin() throws IOException {
// While restoring above, if the HBase Master which was initially the Active one, was down
// and the restore put the cluster back to Initial configuration, HAdmin instance will need
// to refresh its connections (otherwise it will return incorrect information) or we can
@ -303,9 +372,10 @@ public class DistributedHBaseCluster extends HBaseCluster {
try {
admin.close();
} catch (IOException ioe) {
LOG.info("While closing the old connection", ioe);
LOG.warn("While closing the old connection", ioe);
}
this.admin = new HBaseAdmin(conf);
LOG.info("Added new HBaseAdmin");
return true;
}
}

View File

@ -226,16 +226,23 @@ public abstract class HBaseCluster implements Closeable, Configurable {
/**
* Restores the cluster to it's initial state if this is a real cluster,
* otherwise does nothing.
* This is a best effort restore. If the servers are not reachable, or insufficient
* permissions, etc. restoration might be partial.
* @return whether restoration is complete
*/
public void restoreInitialStatus() throws IOException {
restoreClusterStatus(getInitialClusterStatus());
public boolean restoreInitialStatus() throws IOException {
return restoreClusterStatus(getInitialClusterStatus());
}
/**
* Restores the cluster to given state if this is a real cluster,
* otherwise does nothing.
* This is a best effort restore. If the servers are not reachable, or insufficient
* permissions, etc. restoration might be partial.
* @return whether restoration is complete
*/
public void restoreClusterStatus(ClusterStatus desiredStatus) throws IOException {
public boolean restoreClusterStatus(ClusterStatus desiredStatus) throws IOException {
return true;
}
/**