HBASE-17653 HBASE-17624 rsgroup synchronizations will (distributed) deadlock
This patch restores the regime instituted in original rsgroups patch (HBASE-6721) where reading of rsgroup state runs unimpeded against COW immutable Maps whereas mutation to state require exclusive locks (updating the Maps of state when done). HBASE-17624 was over-enthusiastic with its locking down of access making it likely we'd deadlock. Adds documentation on concurrency expectations.
This commit is contained in:
Michael Stack
parent
d7325185ad
commit
b392de3e31
|
|
@ -75,11 +75,15 @@ public class RSGroupAdminServer implements RSGroupAdmin {
|
|||
|
||||
@Override
|
||||
public RSGroupInfo getRSGroupInfoOfTable(TableName tableName) throws IOException {
|
||||
// We are reading across two Maps in the below with out synchronizing across
|
||||
// them; should be safe most of the time.
|
||||
String groupName = getRSGroupInfoManager().getRSGroupOfTable(tableName);
|
||||
return groupName == null? null: getRSGroupInfoManager().getRSGroup(groupName);
|
||||
}
|
||||
|
||||
private void checkOnlineServersOnly(Set<Address> servers) throws ConstraintException {
|
||||
// This uglyness is because we only have Address, not ServerName.
|
||||
// Online servers are keyed by ServerName.
|
||||
Set<Address> onlineServers = new HashSet<Address>();
|
||||
for(ServerName server: master.getServerManager().getOnlineServers().keySet()) {
|
||||
onlineServers.add(server.getAddress());
|
||||
|
|
@ -152,29 +156,29 @@ public class RSGroupAdminServer implements RSGroupAdmin {
|
|||
}
|
||||
RSGroupInfo targetGrp = getAndCheckRSGroupInfo(targetGroupName);
|
||||
RSGroupInfoManager manager = getRSGroupInfoManager();
|
||||
// Lock the manager during the below manipulations.
|
||||
// Hold a lock on the manager instance while moving servers to prevent
|
||||
// another writer changing our state while we are working.
|
||||
synchronized (manager) {
|
||||
if (master.getMasterCoprocessorHost() != null) {
|
||||
master.getMasterCoprocessorHost().preMoveServers(servers, targetGroupName);
|
||||
}
|
||||
// Presume first server is the source group. Later we check all servers are from
|
||||
// this same group.
|
||||
// Presume first server's source group. Later ensure all servers are from this group.
|
||||
Address firstServer = servers.iterator().next();
|
||||
RSGroupInfo srcGrp = manager.getRSGroupOfServer(firstServer);
|
||||
if (srcGrp == null) {
|
||||
// Be careful. This message is tested for in TestRSGroupsBase...
|
||||
// Be careful. This exception message is tested for in TestRSGroupsBase...
|
||||
throw new ConstraintException("Source RSGroup for server " + firstServer + " does not exist.");
|
||||
}
|
||||
if (srcGrp.getName().equals(targetGroupName)) {
|
||||
throw new ConstraintException( "Target RSGroup " + targetGroupName +
|
||||
" is same as source " + srcGrp + " RSGroup.");
|
||||
}
|
||||
// Only move online servers (when from 'default') or servers from other groups.
|
||||
// This prevents bogus servers from entering groups
|
||||
// Only move online servers (when moving from 'default') or servers from other
|
||||
// groups. This prevents bogus servers from entering groups
|
||||
if (RSGroupInfo.DEFAULT_GROUP.equals(srcGrp.getName())) {
|
||||
checkOnlineServersOnly(servers);
|
||||
}
|
||||
// Check all servers are of same rsgroup.
|
||||
// Ensure all servers are of same rsgroup.
|
||||
for (Address server: servers) {
|
||||
String tmpGroup = manager.getRSGroupOfServer(server).getName();
|
||||
if (!tmpGroup.equals(srcGrp.getName())) {
|
||||
|
|
@ -189,8 +193,6 @@ public class RSGroupAdminServer implements RSGroupAdmin {
|
|||
|
||||
// MovedServers may be < passed in 'servers'.
|
||||
Set<Address> movedServers = manager.moveServers(servers, srcGrp.getName(), targetGroupName);
|
||||
// Appy makes note that if we were passed in a List of servers,
|
||||
// we'd save having to do stuff like the below.
|
||||
List<Address> editableMovedServers = Lists.newArrayList(movedServers);
|
||||
boolean foundRegionsToUnassign;
|
||||
do {
|
||||
|
|
@ -202,9 +204,9 @@ public class RSGroupAdminServer implements RSGroupAdmin {
|
|||
|
||||
// Unassign regions for a server
|
||||
// TODO: This is problematic especially if hbase:meta is in the mix.
|
||||
// We need to update state in hbase:meta and if unassigned we hang
|
||||
// We need to update state in hbase:meta on Master and if unassigned we hang
|
||||
// around in here. There is a silly sort on linked list done above
|
||||
// in getRegions putting hbase:meta last which helps but probably holes.
|
||||
// in getRegions putting hbase:meta last which helps but probably has holes.
|
||||
LOG.info("Unassigning " + regions.size() +
|
||||
" region(s) from " + rs + " for server move to " + targetGroupName);
|
||||
if (!regions.isEmpty()) {
|
||||
|
|
@ -253,12 +255,12 @@ public class RSGroupAdminServer implements RSGroupAdmin {
|
|||
return;
|
||||
}
|
||||
RSGroupInfoManager manager = getRSGroupInfoManager();
|
||||
// Lock the manager during below machinations.
|
||||
// Hold a lock on the manager instance while moving servers to prevent
|
||||
// another writer changing our state while we are working.
|
||||
synchronized (manager) {
|
||||
if (master.getMasterCoprocessorHost() != null) {
|
||||
master.getMasterCoprocessorHost().preMoveTables(tables, targetGroup);
|
||||
}
|
||||
|
||||
if(targetGroup != null) {
|
||||
RSGroupInfo destGroup = manager.getRSGroup(targetGroup);
|
||||
if(destGroup == null) {
|
||||
|
|
@ -315,22 +317,23 @@ public class RSGroupAdminServer implements RSGroupAdmin {
|
|||
@Override
|
||||
public void removeRSGroup(String name) throws IOException {
|
||||
RSGroupInfoManager manager = getRSGroupInfoManager();
|
||||
// Hold lock across coprocessor calls.
|
||||
// Hold a lock on the manager instance while moving servers to prevent
|
||||
// another writer changing our state while we are working.
|
||||
synchronized (manager) {
|
||||
if (master.getMasterCoprocessorHost() != null) {
|
||||
master.getMasterCoprocessorHost().preRemoveRSGroup(name);
|
||||
}
|
||||
RSGroupInfo RSGroupInfo = manager.getRSGroup(name);
|
||||
if (RSGroupInfo == null) {
|
||||
RSGroupInfo rsgi = manager.getRSGroup(name);
|
||||
if (rsgi == null) {
|
||||
throw new ConstraintException("RSGroup " + name + " does not exist");
|
||||
}
|
||||
int tableCount = RSGroupInfo.getTables().size();
|
||||
int tableCount = rsgi.getTables().size();
|
||||
if (tableCount > 0) {
|
||||
throw new ConstraintException("RSGroup " + name + " has " + tableCount +
|
||||
" tables; you must remove these tables from the rsgroup before " +
|
||||
"the rsgroup can be removed.");
|
||||
}
|
||||
int serverCount = RSGroupInfo.getServers().size();
|
||||
int serverCount = rsgi.getServers().size();
|
||||
if (serverCount > 0) {
|
||||
throw new ConstraintException("RSGroup " + name + " has " + serverCount +
|
||||
" servers; you must remove these servers from the RSGroup before" +
|
||||
|
|
@ -465,7 +468,7 @@ public class RSGroupAdminServer implements RSGroupAdmin {
|
|||
Map<ServerName, List<HRegionInfo>> serverMap = Maps.newHashMap();
|
||||
for(ServerName serverName: master.getServerManager().getOnlineServers().keySet()) {
|
||||
if(RSGroupInfo.getServers().contains(serverName.getAddress())) {
|
||||
serverMap.put(serverName, Collections.EMPTY_LIST);
|
||||
serverMap.put(serverName, Collections.emptyList());
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -79,15 +79,12 @@ public class RSGroupBasedLoadBalancer implements RSGroupableBalancer, LoadBalanc
|
|||
private Configuration config;
|
||||
private ClusterStatus clusterStatus;
|
||||
private MasterServices masterServices;
|
||||
// Synchronize on access until we take the time to cmoe up with a finer-grained
|
||||
// locking regime.
|
||||
private volatile RSGroupInfoManager rsGroupInfoManager;
|
||||
private LoadBalancer internalBalancer;
|
||||
|
||||
//used during reflection by LoadBalancerFactory
|
||||
@InterfaceAudience.Private
|
||||
public RSGroupBasedLoadBalancer() {
|
||||
}
|
||||
public RSGroupBasedLoadBalancer() {}
|
||||
|
||||
//This constructor should only be used for unit testing
|
||||
@InterfaceAudience.Private
|
||||
|
|
@ -323,9 +320,7 @@ public class RSGroupBasedLoadBalancer implements RSGroupableBalancer, LoadBalanc
|
|||
if (assignedServer != null &&
|
||||
(info == null || !info.containsServer(assignedServer.getAddress()))) {
|
||||
RSGroupInfo otherInfo = null;
|
||||
synchronized (this.rsGroupInfoManager) {
|
||||
otherInfo = rsGroupInfoManager.getRSGroupOfServer(assignedServer.getAddress());
|
||||
}
|
||||
otherInfo = rsGroupInfoManager.getRSGroupOfServer(assignedServer.getAddress());
|
||||
LOG.debug("Found misplaced region: " + regionInfo.getRegionNameAsString() +
|
||||
" on server: " + assignedServer +
|
||||
" found in group: " + otherInfo +
|
||||
|
|
|
|||
|
|
@ -91,16 +91,22 @@ import com.google.protobuf.ServiceException;
|
|||
* for bootstrapping during offline mode.
|
||||
*
|
||||
* <h2>Concurrency</h2>
|
||||
* All methods are synchronized to protect against concurrent access on contained
|
||||
* Maps and so as only one writer at a time to the backing zookeeper cache and rsgroup table.
|
||||
* RSGroup state is kept locally in Maps. There is a rsgroup name to cached
|
||||
* RSGroupInfo Map at this.rsGroupMap and a Map of tables to the name of the
|
||||
* rsgroup they belong too (in this.tableMap). These Maps are persisted to the
|
||||
* hbase:rsgroup table (and cached in zk) on each modification.
|
||||
*
|
||||
* <p>Mutations on state are synchronized but so reads can continue without having
|
||||
* to wait on an instance monitor, mutations do wholesale replace of the Maps on
|
||||
* update -- Copy-On-Write; the local Maps of state are read-only, just-in-case
|
||||
* (see flushConfig).
|
||||
*
|
||||
* <p>Reads must not block else there is a danger we'll deadlock.
|
||||
*
|
||||
* <p>Clients of this class, the {@link RSGroupAdminEndpoint} for example, want to query and
|
||||
* then act on the results of the query modifying cache in zookeeper without another thread
|
||||
* making intermediate modifications. These clients synchronize on the 'this' instance so
|
||||
* no other has access concurrently.
|
||||
*
|
||||
* TODO: Spend time cleaning up this coarse locking that is prone to error if not carefully
|
||||
* enforced everywhere.
|
||||
* no other has access concurrently. Reads must be able to continue concurrently.
|
||||
*/
|
||||
@InterfaceAudience.Private
|
||||
public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListener {
|
||||
|
|
@ -121,38 +127,37 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
}
|
||||
}
|
||||
|
||||
private Map<String, RSGroupInfo> rsGroupMap;
|
||||
private Map<TableName, String> tableMap;
|
||||
// There two Maps are immutable and wholesale replaced on each modification
|
||||
// so are safe to access concurrently. See class comment.
|
||||
private volatile Map<String, RSGroupInfo> rsGroupMap = Collections.emptyMap();
|
||||
private volatile Map<TableName, String> tableMap = Collections.emptyMap();
|
||||
|
||||
private final MasterServices master;
|
||||
private Table rsGroupTable;
|
||||
private final ClusterConnection conn;
|
||||
private final ZooKeeperWatcher watcher;
|
||||
private RSGroupStartupWorker rsGroupStartupWorker;
|
||||
// contains list of groups that were last flushed to persistent store
|
||||
private Set<String> prevRSGroups;
|
||||
private final RSGroupSerDe rsGroupSerDe;
|
||||
private Set<String> prevRSGroups = new HashSet<String>();
|
||||
private final RSGroupSerDe rsGroupSerDe = new RSGroupSerDe();
|
||||
private DefaultServerUpdater defaultServerUpdater;
|
||||
private boolean init = false;
|
||||
|
||||
|
||||
public RSGroupInfoManagerImpl(MasterServices master) throws IOException {
|
||||
this.rsGroupMap = Collections.emptyMap();
|
||||
this.tableMap = Collections.emptyMap();
|
||||
rsGroupSerDe = new RSGroupSerDe();
|
||||
this.master = master;
|
||||
this.watcher = master.getZooKeeper();
|
||||
this.conn = master.getClusterConnection();
|
||||
prevRSGroups = new HashSet<String>();
|
||||
}
|
||||
|
||||
public synchronized void init() throws IOException{
|
||||
if (this.init) return;
|
||||
rsGroupStartupWorker = new RSGroupStartupWorker(this, master, conn);
|
||||
refresh();
|
||||
rsGroupStartupWorker.start();
|
||||
defaultServerUpdater = new DefaultServerUpdater(this);
|
||||
master.getServerManager().registerListener(this);
|
||||
defaultServerUpdater.start();
|
||||
init = true;
|
||||
this.init = true;
|
||||
}
|
||||
|
||||
synchronized boolean isInit() {
|
||||
|
|
@ -196,14 +201,13 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
RSGroupInfo src = getRSGroupInfo(srcGroup);
|
||||
RSGroupInfo dst = getRSGroupInfo(dstGroup);
|
||||
// If destination is 'default' rsgroup, only add servers that are online. If not online, drop it.
|
||||
// If not 'default' group, add server to dst group EVEN IF IT IS NOT online (could be a group
|
||||
// If not 'default' group, add server to 'dst' rsgroup EVEN IF IT IS NOT online (could be a rsgroup
|
||||
// of dead servers that are to come back later).
|
||||
Set<Address> onlineServers = dst.getName().equals(RSGroupInfo.DEFAULT_GROUP)?
|
||||
getOnlineServers(this.master): null;
|
||||
for (Address el: servers) {
|
||||
src.removeServer(el);
|
||||
if (onlineServers != null) {
|
||||
// onlineServers is non-null if 'default' rsgroup. If the server is not online, drop it.
|
||||
if (!onlineServers.contains(el)) {
|
||||
if (LOG.isDebugEnabled()) {
|
||||
LOG.debug("Dropping " + el + " during move-to-default rsgroup because not online");
|
||||
|
|
@ -227,9 +231,9 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
* @return An instance of GroupInfo.
|
||||
*/
|
||||
@Override
|
||||
public synchronized RSGroupInfo getRSGroupOfServer(Address hostPort)
|
||||
public RSGroupInfo getRSGroupOfServer(Address hostPort)
|
||||
throws IOException {
|
||||
for (RSGroupInfo info : rsGroupMap.values()) {
|
||||
for (RSGroupInfo info: rsGroupMap.values()) {
|
||||
if (info.containsServer(hostPort)) {
|
||||
return info;
|
||||
}
|
||||
|
|
@ -245,14 +249,14 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
* @return An instance of GroupInfo
|
||||
*/
|
||||
@Override
|
||||
public synchronized RSGroupInfo getRSGroup(String groupName) throws IOException {
|
||||
public RSGroupInfo getRSGroup(String groupName) throws IOException {
|
||||
return this.rsGroupMap.get(groupName);
|
||||
}
|
||||
|
||||
|
||||
|
||||
@Override
|
||||
public synchronized String getRSGroupOfTable(TableName tableName) throws IOException {
|
||||
public String getRSGroupOfTable(TableName tableName) throws IOException {
|
||||
return tableMap.get(tableName);
|
||||
}
|
||||
|
||||
|
|
@ -298,12 +302,12 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
}
|
||||
|
||||
@Override
|
||||
public synchronized List<RSGroupInfo> listRSGroups() throws IOException {
|
||||
public List<RSGroupInfo> listRSGroups() throws IOException {
|
||||
return Lists.newLinkedList(rsGroupMap.values());
|
||||
}
|
||||
|
||||
@Override
|
||||
public synchronized boolean isOnline() {
|
||||
public boolean isOnline() {
|
||||
return rsGroupStartupWorker.isOnline();
|
||||
}
|
||||
|
||||
|
|
@ -312,10 +316,16 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
refresh(false);
|
||||
}
|
||||
|
||||
/**
|
||||
* Read rsgroup info from the source of truth, the hbase:rsgroup table.
|
||||
* Update zk cache. Called on startup of the manager.
|
||||
* @param forceOnline
|
||||
* @throws IOException
|
||||
*/
|
||||
private synchronized void refresh(boolean forceOnline) throws IOException {
|
||||
List<RSGroupInfo> groupList = new LinkedList<RSGroupInfo>();
|
||||
|
||||
// overwrite anything read from zk, group table is source of truth
|
||||
// Overwrite anything read from zk, group table is source of truth
|
||||
// if online read from GROUP table
|
||||
if (forceOnline || isOnline()) {
|
||||
LOG.debug("Refreshing in Online mode.");
|
||||
|
|
@ -324,7 +334,7 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
}
|
||||
groupList.addAll(rsGroupSerDe.retrieveGroupList(rsGroupTable));
|
||||
} else {
|
||||
LOG.debug("Refershing in Offline mode.");
|
||||
LOG.debug("Refreshing in Offline mode.");
|
||||
String groupBasePath = ZKUtil.joinZNode(watcher.znodePaths.baseZNode, rsGroupZNode);
|
||||
groupList.addAll(rsGroupSerDe.retrieveGroupList(watcher, groupBasePath));
|
||||
}
|
||||
|
|
@ -347,20 +357,18 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
master.listTableNamesByNamespace(NamespaceDescriptor.SYSTEM_NAMESPACE_NAME_STR);
|
||||
}
|
||||
|
||||
for(TableName table : specialTables) {
|
||||
for (TableName table : specialTables) {
|
||||
orphanTables.add(table);
|
||||
}
|
||||
for(RSGroupInfo group: groupList) {
|
||||
for (RSGroupInfo group: groupList) {
|
||||
if(!group.getName().equals(RSGroupInfo.DEFAULT_GROUP)) {
|
||||
orphanTables.removeAll(group.getTables());
|
||||
}
|
||||
}
|
||||
|
||||
// This is added to the last of the list
|
||||
// so it overwrites the default group loaded
|
||||
// This is added to the last of the list so it overwrites the 'default' rsgroup loaded
|
||||
// from region group table or zk
|
||||
groupList.add(new RSGroupInfo(RSGroupInfo.DEFAULT_GROUP,
|
||||
Sets.newTreeSet(getDefaultServers()),
|
||||
groupList.add(new RSGroupInfo(RSGroupInfo.DEFAULT_GROUP, getDefaultServers(),
|
||||
orphanTables));
|
||||
|
||||
// populate the data
|
||||
|
|
@ -372,11 +380,8 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
newTableMap.put(table, group.getName());
|
||||
}
|
||||
}
|
||||
rsGroupMap = Collections.unmodifiableMap(newGroupMap);
|
||||
tableMap = Collections.unmodifiableMap(newTableMap);
|
||||
|
||||
prevRSGroups.clear();
|
||||
prevRSGroups.addAll(rsGroupMap.keySet());
|
||||
installNewMaps(newGroupMap, newTableMap);
|
||||
updateCacheOfRSGroups(rsGroupMap.keySet());
|
||||
}
|
||||
|
||||
private synchronized Map<TableName,String> flushConfigTable(Map<String,RSGroupInfo> newGroupMap)
|
||||
|
|
@ -408,12 +413,14 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
}
|
||||
return newTableMap;
|
||||
}
|
||||
private synchronized void flushConfig() throws IOException {
|
||||
flushConfig(rsGroupMap);
|
||||
|
||||
private synchronized void flushConfig()
|
||||
throws IOException {
|
||||
flushConfig(this.rsGroupMap);
|
||||
}
|
||||
|
||||
// Called from RSGroupStartupWorker thread so synchronize
|
||||
private synchronized void flushConfig(Map<String, RSGroupInfo> newGroupMap) throws IOException {
|
||||
private synchronized void flushConfig(Map<String, RSGroupInfo> newGroupMap)
|
||||
throws IOException {
|
||||
Map<TableName, String> newTableMap;
|
||||
|
||||
// For offline mode persistence is still unavailable
|
||||
|
|
@ -433,11 +440,8 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
|
||||
newTableMap = flushConfigTable(newGroupMap);
|
||||
|
||||
// make changes visible since it has been
|
||||
// persisted in the source of truth
|
||||
rsGroupMap = Collections.unmodifiableMap(newGroupMap);
|
||||
tableMap = Collections.unmodifiableMap(newTableMap);
|
||||
|
||||
// Make changes visible after having been persisted to the source of truth
|
||||
installNewMaps(newGroupMap, newTableMap);
|
||||
|
||||
try {
|
||||
String groupBasePath = ZKUtil.joinZNode(watcher.znodePaths.baseZNode, rsGroupZNode);
|
||||
|
|
@ -469,9 +473,28 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
master.abort("Failed to write to rsGroupZNode", e);
|
||||
throw new IOException("Failed to write to rsGroupZNode",e);
|
||||
}
|
||||
updateCacheOfRSGroups(newGroupMap.keySet());
|
||||
}
|
||||
|
||||
prevRSGroups.clear();
|
||||
prevRSGroups.addAll(newGroupMap.keySet());
|
||||
/**
|
||||
* Make changes visible.
|
||||
* Caller must be synchronized on 'this'.
|
||||
*/
|
||||
private void installNewMaps(Map<String, RSGroupInfo> newRSGroupMap,
|
||||
Map<TableName, String> newTableMap) {
|
||||
// Make maps Immutable.
|
||||
this.rsGroupMap = Collections.unmodifiableMap(newRSGroupMap);
|
||||
this.tableMap = Collections.unmodifiableMap(newTableMap);
|
||||
}
|
||||
|
||||
/**
|
||||
* Update cache of rsgroups.
|
||||
* Caller must be synchronized on 'this'.
|
||||
* @param currentGroups Current list of Groups.
|
||||
*/
|
||||
private void updateCacheOfRSGroups(final Set<String> currentGroups) {
|
||||
this.prevRSGroups.clear();
|
||||
this.prevRSGroups.addAll(currentGroups);
|
||||
}
|
||||
|
||||
// Called by getDefaultServers. Presume it has lock in place.
|
||||
|
|
@ -494,18 +517,19 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
// Called by DefaultServerUpdater. Presume it has lock on this manager when it runs.
|
||||
private SortedSet<Address> getDefaultServers() throws IOException {
|
||||
SortedSet<Address> defaultServers = Sets.newTreeSet();
|
||||
for (ServerName server : getOnlineRS()) {
|
||||
Address hostPort = Address.fromParts(server.getHostname(), server.getPort());
|
||||
for (ServerName serverName : getOnlineRS()) {
|
||||
Address server =
|
||||
Address.fromParts(serverName.getHostname(), serverName.getPort());
|
||||
boolean found = false;
|
||||
for(RSGroupInfo RSGroupInfo: listRSGroups()) {
|
||||
if(!RSGroupInfo.DEFAULT_GROUP.equals(RSGroupInfo.getName()) &&
|
||||
RSGroupInfo.containsServer(hostPort)) {
|
||||
for(RSGroupInfo rsgi: listRSGroups()) {
|
||||
if(!RSGroupInfo.DEFAULT_GROUP.equals(rsgi.getName()) &&
|
||||
rsgi.containsServer(server)) {
|
||||
found = true;
|
||||
break;
|
||||
}
|
||||
}
|
||||
if (!found) {
|
||||
defaultServers.add(hostPort);
|
||||
defaultServers.add(server);
|
||||
}
|
||||
}
|
||||
return defaultServers;
|
||||
|
|
@ -533,10 +557,12 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
defaultServerUpdater.serverChanged();
|
||||
}
|
||||
|
||||
// TODO: Why do we need this extra thread? Why can't we just go
|
||||
// fetch at balance time or admin time?
|
||||
private static class DefaultServerUpdater extends Thread {
|
||||
private static final Log LOG = LogFactory.getLog(DefaultServerUpdater.class);
|
||||
private final RSGroupInfoManagerImpl mgr;
|
||||
private boolean hasChanged = false;
|
||||
private boolean changed = false;
|
||||
|
||||
public DefaultServerUpdater(RSGroupInfoManagerImpl mgr) {
|
||||
super("RSGroup.ServerUpdater");
|
||||
|
|
@ -550,20 +576,18 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
while(isMasterRunning(this.mgr.master)) {
|
||||
try {
|
||||
LOG.info("Updating default servers.");
|
||||
synchronized (this.mgr) {
|
||||
SortedSet<Address> servers = mgr.getDefaultServers();
|
||||
if (!servers.equals(prevDefaultServers)) {
|
||||
mgr.updateDefaultServers(servers);
|
||||
prevDefaultServers = servers;
|
||||
LOG.info("Updated with servers: "+servers.size());
|
||||
}
|
||||
SortedSet<Address> servers = mgr.getDefaultServers();
|
||||
if (!servers.equals(prevDefaultServers)) {
|
||||
mgr.updateDefaultServers(servers);
|
||||
prevDefaultServers = servers;
|
||||
LOG.info("Updated with servers: " + servers.size());
|
||||
}
|
||||
try {
|
||||
synchronized (this) {
|
||||
if(!hasChanged) {
|
||||
if(!changed) {
|
||||
wait();
|
||||
}
|
||||
hasChanged = false;
|
||||
changed = false;
|
||||
}
|
||||
} catch (InterruptedException e) {
|
||||
LOG.warn("Interrupted", e);
|
||||
|
|
@ -576,7 +600,7 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
|
||||
public void serverChanged() {
|
||||
synchronized (this) {
|
||||
hasChanged = true;
|
||||
changed = true;
|
||||
this.notify();
|
||||
}
|
||||
}
|
||||
|
|
@ -736,7 +760,7 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
}
|
||||
tries--;
|
||||
}
|
||||
if(tries <= 0) {
|
||||
if (tries <= 0) {
|
||||
throw new IOException("Failed to create group table in a given time.");
|
||||
} else {
|
||||
ProcedureInfo result = masterServices.getMasterProcedureExecutor().getResult(procId);
|
||||
|
|
@ -777,8 +801,8 @@ public class RSGroupInfoManagerImpl implements RSGroupInfoManager, ServerListene
|
|||
}
|
||||
|
||||
private void checkGroupName(String groupName) throws ConstraintException {
|
||||
if(!groupName.matches("[a-zA-Z0-9_]+")) {
|
||||
throw new ConstraintException("Group name should only contain alphanumeric characters");
|
||||
if (!groupName.matches("[a-zA-Z0-9_]+")) {
|
||||
throw new ConstraintException("RSGroup name should only contain alphanumeric characters");
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -60,7 +60,7 @@ public class TestRSGroups extends TestRSGroupsBase {
|
|||
protected static final Log LOG = LogFactory.getLog(TestRSGroups.class);
|
||||
private static HMaster master;
|
||||
private static boolean INIT = false;
|
||||
private static RSGroupAdminEndpoint RSGroupAdminEndpoint;
|
||||
private static RSGroupAdminEndpoint rsGroupAdminEndpoint;
|
||||
|
||||
|
||||
@BeforeClass
|
||||
|
|
@ -93,7 +93,7 @@ public class TestRSGroups extends TestRSGroupsBase {
|
|||
admin.setBalancerRunning(false,true);
|
||||
rsGroupAdmin = new VerifyingRSGroupAdminClient(
|
||||
new RSGroupAdminClient(TEST_UTIL.getConnection()), TEST_UTIL.getConfiguration());
|
||||
RSGroupAdminEndpoint =
|
||||
rsGroupAdminEndpoint =
|
||||
master.getMasterCoprocessorHost().findCoprocessors(RSGroupAdminEndpoint.class).get(0);
|
||||
}
|
||||
|
||||
|
|
@ -236,11 +236,8 @@ public class TestRSGroups extends TestRSGroupsBase {
|
|||
|
||||
@Test
|
||||
public void testGroupInfoMultiAccessing() throws Exception {
|
||||
RSGroupInfoManager manager = RSGroupAdminEndpoint.getGroupInfoManager();
|
||||
RSGroupInfo defaultGroup = null;
|
||||
synchronized (manager) {
|
||||
defaultGroup = manager.getRSGroup("default");
|
||||
}
|
||||
RSGroupInfoManager manager = rsGroupAdminEndpoint.getGroupInfoManager();
|
||||
RSGroupInfo defaultGroup = manager.getRSGroup("default");
|
||||
// getRSGroup updates default group's server list
|
||||
// this process must not affect other threads iterating the list
|
||||
Iterator<Address> it = defaultGroup.getServers().iterator();
|
||||
|
|
@ -258,7 +255,7 @@ public class TestRSGroups extends TestRSGroupsBase {
|
|||
TEST_UTIL.createMultiRegionTable(tableName, new byte[]{'f'}, 15);
|
||||
TEST_UTIL.waitUntilAllRegionsAssigned(tableName);
|
||||
|
||||
RSGroupAdminEndpoint.getGroupInfoManager()
|
||||
rsGroupAdminEndpoint.getGroupInfoManager()
|
||||
.moveTables(Sets.newHashSet(tableName), RSGroupInfo.getName());
|
||||
|
||||
assertTrue(rsGroupAdmin.balanceRSGroup(RSGroupInfo.getName()));
|
||||
|
|
|
|||
|
|
@ -162,13 +162,11 @@ public class TestRSGroupsOfflineMode {
|
|||
|
||||
RSGroupInfoManager groupMgr = RSGroupAdminEndpoint.getGroupInfoManager();
|
||||
//make sure balancer is in offline mode, since this is what we're testing
|
||||
synchronized (groupMgr) {
|
||||
assertFalse(groupMgr.isOnline());
|
||||
//verify the group affiliation that's loaded from ZK instead of tables
|
||||
assertEquals(newGroup,
|
||||
assertFalse(groupMgr.isOnline());
|
||||
//verify the group affiliation that's loaded from ZK instead of tables
|
||||
assertEquals(newGroup,
|
||||
groupMgr.getRSGroupOfTable(RSGroupInfoManager.RSGROUP_TABLE_NAME));
|
||||
assertEquals(RSGroupInfo.DEFAULT_GROUP, groupMgr.getRSGroupOfTable(failoverTable));
|
||||
}
|
||||
assertEquals(RSGroupInfo.DEFAULT_GROUP, groupMgr.getRSGroupOfTable(failoverTable));
|
||||
//kill final regionserver to see the failover happens for all tables
|
||||
//except GROUP table since it's group does not have any online RS
|
||||
killRS.stop("die");
|
||||
|
|
|
|||
Loading…
Reference in New Issue