org.apache.solr.cloud

Class ZkController

java.lang.Object

org.apache.solr.cloud.ZkController

public class ZkController
extends Object

Handle ZooKeeper interactions.

notes: loads everything on init, creates what's not there - further updates are prompted with Watches.

TODO: exceptions during close on attempts to update cloud state

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	ZkController.NotInClusterStateException Thrown during pre register process if the replica is not present in clusterstate
static class	ZkController.NotLeaderException Thrown during leader initiated recovery process if current node is not leader
static class	ZkController.ResourceModifiedInZkException

Field Summary

Fields

Modifier and Type	Field and Description
static String	COLLECTION_PARAM_PREFIX
static String	CONFIGNAME_PROP
protected Overseer	overseer
ZkStateReader	zkStateReader

Constructor Summary

Constructors

Constructor and Description
ZkController(CoreContainer cc, String zkServerAddress, int zkClientConnectTimeout, CloudConfig cloudConfig, CurrentCoreDescriptorProvider registerOnReconnect)

Method Summary

Modifier and Type	Method and Description
void	addOnReconnectListener(OnReconnect listener) Add a listener to be notified once there is a new session created after a ZooKeeper session expiration occurs; in most cases, listeners will be components that have watchers that need to be re-created.
static void	bootstrapConf(SolrZkClient zkClient, CoreContainer cc, String solrHome) If in SolrCloud mode, upload config sets for each SolrCore in solr.xml.
static boolean	checkChrootPath(String zkHost, boolean create) Validates if the chroot exists in zk (or if it is successfully created).
boolean	checkIfCoreNodeNameAlreadyExists(CoreDescriptor dcore)
void	checkOverseerDesignate()
boolean	claimAsyncId(String asyncId) When an operation needs to be performed in an asynchronous mode, the asyncId needs to be claimed by calling this method to make sure it's not duplicate (hasn't been claimed by other request).
boolean	clearAsyncId(String asyncId) Clears an asyncId previously claimed by calling claimAsyncId(String)
void	clearZkCollectionTerms()
void	close() Closes the underlying ZooKeeper client.
boolean	configFileExists(String collection, String fileName) Returns true if config file exists
static void	createClusterZkNodes(SolrZkClient zkClient) Create the zknodes necessary for a cluster to operate
void	createCollection(String collection)
boolean	ensureReplicaInLeaderInitiatedRecovery(CoreContainer container, String collection, String shardId, ZkCoreNodeProps replicaCoreProps, CoreDescriptor leaderCd, boolean forcePublishState) Deprecated.
String	getBaseUrl()
int	getClientTimeout()
ClusterState	getClusterState()
OnReconnect	getConfigDirListener()
byte[]	getConfigFileData(String zkConfigName, String fileName) Returns config file data (in bytes)
CoreContainer	getCoreContainer()
String	getCoreNodeName(CoreDescriptor descriptor)
String	getHostName()
int	getHostPort()
int	getLeaderConflictResolveWait()
Replica.State	getLeaderInitiatedRecoveryState(String collection, String shardId, String coreNodeName) Deprecated.
Map<String,Object>	getLeaderInitiatedRecoveryStateObject(String collection, String shardId, String coreNodeName) Deprecated.
static String	getLeaderInitiatedRecoveryZnodePath(String collection, String shardId) Deprecated.
static String	getLeaderInitiatedRecoveryZnodePath(String collection, String shardId, String coreNodeName) Deprecated.
ZkCoreNodeProps	getLeaderProps(String collection, String slice, int timeoutms) Get leader props directly from zk nodes.
ZkCoreNodeProps	getLeaderProps(String collection, String slice, int timeoutms, boolean failImmediatelyOnExpiration) Get leader props directly from zk nodes.
int	getLeaderVoteWait()
String	getNodeName()
Overseer	getOverseer()
OverseerTaskQueue	getOverseerCollectionQueue()
DistributedMap	getOverseerCompletedMap()
OverseerTaskQueue	getOverseerConfigSetQueue()
LeaderElector	getOverseerElector()
DistributedMap	getOverseerFailureMap()
ZkDistributedQueue	getOverseerJobQueue()
DistributedMap	getOverseerRunningMap()
ZkShardTerms	getShardTerms(String collection, String shardId)
SolrCloudManager	getSolrCloudManager()
SolrZkClient	getZkClient()
String	getZkServerAddress()
ZkStateReader	getZkStateReader()
void	giveupLeadership(CoreDescriptor cd, Throwable tragicException)
boolean	isConnected()
boolean	isReplicaInRecoveryHandling(String replicaUrl) Deprecated.
static void	linkConfSet(SolrZkClient zkClient, String collection, String confSetName)
boolean	pathExists(String path) Returns true if the path exists
static int	persistConfigResourceToZooKeeper(ZkSolrResourceLoader zkLoader, int znodeVersion, String resourceName, byte[] content, boolean createIfNotExists) Persists a config file to ZooKeeper using optimistic concurrency.
void	preRegister(CoreDescriptor cd, boolean publishState)
void	publish(CoreDescriptor cd, Replica.State state)
void	publish(CoreDescriptor cd, Replica.State state, boolean updateLastState, boolean forcePublish) Publish core state to overseer.
void	publishAndWaitForDownStates()
void	publishNodeAsDown(String nodeName) Best effort to set DOWN state for all replicas on node.
String	register(String coreName, CoreDescriptor desc, boolean skipRecovery) Register shard with ZooKeeper.
String	register(String coreName, CoreDescriptor desc, boolean recoverReloadedCores, boolean afterExpiration, boolean skipRecovery) Register shard with ZooKeeper.
void	registerConfListenerForCore(String confDir, SolrCore core, Runnable listener) This will give a callback to the listener whenever a child is modified in the conf directory.
void	rejoinOverseerElection(String electionNode, boolean joinAtHead)
void	rejoinShardLeaderElection(SolrParams params)
void	removeEphemeralLiveNode()
void	removeOnReconnectListener(OnReconnect listener) Removed a previously registered OnReconnect listener, such as when a core is removed or reloaded.
void	removeReplicaFromLeaderInitiatedRecoveryHandling(String replicaUrl) Deprecated.
void	startReplicationFromLeader(String coreName, boolean switchTransactionLog)
void	stopReplicationFromLeader(String coreName)
void	throwErrorIfReplicaReplaced(CoreDescriptor desc)
static void	touchConfDir(ZkSolrResourceLoader zkLoader)
static String	trimLeadingAndTrailingSlashes(String in) Utility method for trimming and leading and/or trailing slashes from its input.
void	unregister(String coreName, CoreDescriptor cd)
void	unregister(String coreName, CoreDescriptor cd, boolean removeCoreFromZk)
void	updateLeaderInitiatedRecoveryState(String collection, String shardId, String coreNodeName, Replica.State state, CoreDescriptor leaderCd, boolean retryOnConnLoss) Deprecated.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

COLLECTION_PARAM_PREFIX

public static final String COLLECTION_PARAM_PREFIX

See Also:

Constant Field Values

CONFIGNAME_PROP

public static final String CONFIGNAME_PROP

See Also:

Constant Field Values

zkStateReader

public final ZkStateReader zkStateReader

overseer

protected volatile Overseer overseer

Constructor Detail

ZkController

public ZkController(CoreContainer cc,
                    String zkServerAddress,
                    int zkClientConnectTimeout,
                    CloudConfig cloudConfig,
                    CurrentCoreDescriptorProvider registerOnReconnect)
             throws InterruptedException,
                    TimeoutException,
                    IOException

Throws:

InterruptedException

TimeoutException

IOException

Method Detail

getLeaderVoteWait

public int getLeaderVoteWait()

getLeaderConflictResolveWait

public int getLeaderConflictResolveWait()

close

public void close()

Closes the underlying ZooKeeper client.

giveupLeadership

public void giveupLeadership(CoreDescriptor cd,
                             Throwable tragicException)

configFileExists

public boolean configFileExists(String collection,
                                String fileName)
                         throws org.apache.zookeeper.KeeperException,
                                InterruptedException

Returns true if config file exists

Throws:

org.apache.zookeeper.KeeperException

InterruptedException

getClusterState

public ClusterState getClusterState()

Returns:

information about the cluster from ZooKeeper

getSolrCloudManager

public SolrCloudManager getSolrCloudManager()

getConfigFileData

public byte[] getConfigFileData(String zkConfigName,
                                String fileName)
                         throws org.apache.zookeeper.KeeperException,
                                InterruptedException

Returns config file data (in bytes)

Throws:

org.apache.zookeeper.KeeperException

InterruptedException

getHostName

public String getHostName()

getHostPort

public int getHostPort()

getZkClient

public SolrZkClient getZkClient()

getZkServerAddress

public String getZkServerAddress()

Returns:

zookeeper server address

createClusterZkNodes

public static void createClusterZkNodes(SolrZkClient zkClient)
                                 throws org.apache.zookeeper.KeeperException,
                                        InterruptedException,
                                        IOException

Create the zknodes necessary for a cluster to operate

Parameters:

zkClient - a SolrZkClient

Throws:

org.apache.zookeeper.KeeperException - if there is a Zookeeper error

InterruptedException - on interrupt

IOException

publishAndWaitForDownStates

public void publishAndWaitForDownStates()
                                 throws org.apache.zookeeper.KeeperException,
                                        InterruptedException

Throws:

org.apache.zookeeper.KeeperException

InterruptedException

checkChrootPath

public static boolean checkChrootPath(String zkHost,
                                      boolean create)
                               throws org.apache.zookeeper.KeeperException,
                                      InterruptedException

Validates if the chroot exists in zk (or if it is successfully created). Optionally, if create is set to true this method will create the path in case it doesn't exist

Returns:

true if the path exists or is created false if the path doesn't exist and 'create' = false

Throws:

org.apache.zookeeper.KeeperException

InterruptedException

isConnected

public boolean isConnected()

removeEphemeralLiveNode

public void removeEphemeralLiveNode()
                             throws org.apache.zookeeper.KeeperException,
                                    InterruptedException

Throws:

org.apache.zookeeper.KeeperException

InterruptedException

getNodeName

public String getNodeName()

pathExists

public boolean pathExists(String path)
                   throws org.apache.zookeeper.KeeperException,
                          InterruptedException

Returns true if the path exists

Throws:

org.apache.zookeeper.KeeperException

InterruptedException

register

public String register(String coreName,
                       CoreDescriptor desc,
                       boolean skipRecovery)
                throws Exception

Register shard with ZooKeeper.

Returns:

the shardId for the SolrCore

Throws:

Exception

register

public String register(String coreName,
                       CoreDescriptor desc,
                       boolean recoverReloadedCores,
                       boolean afterExpiration,
                       boolean skipRecovery)
                throws Exception

Register shard with ZooKeeper.

Returns:

the shardId for the SolrCore

Throws:

Exception

startReplicationFromLeader

public void startReplicationFromLeader(String coreName,
                                       boolean switchTransactionLog)
                                throws InterruptedException

Throws:

InterruptedException

stopReplicationFromLeader

public void stopReplicationFromLeader(String coreName)

getLeaderProps

public ZkCoreNodeProps getLeaderProps(String collection,
                                      String slice,
                                      int timeoutms)
                               throws InterruptedException

Get leader props directly from zk nodes.

Throws:

InterruptedException

getLeaderProps

public ZkCoreNodeProps getLeaderProps(String collection,
                                      String slice,
                                      int timeoutms,
                                      boolean failImmediatelyOnExpiration)
                               throws InterruptedException

Get leader props directly from zk nodes.

Returns:

leader props

Throws:

InterruptedException

getBaseUrl

public String getBaseUrl()

publish

public void publish(CoreDescriptor cd,
                    Replica.State state)
             throws Exception

Throws:

Exception

publish

public void publish(CoreDescriptor cd,
                    Replica.State state,
                    boolean updateLastState,
                    boolean forcePublish)
             throws Exception

Publish core state to overseer.

Throws:

Exception

getShardTerms

public ZkShardTerms getShardTerms(String collection,
                                  String shardId)

clearZkCollectionTerms

public void clearZkCollectionTerms()

unregister

public void unregister(String coreName,
                       CoreDescriptor cd)
                throws Exception

Throws:

Exception

unregister

public void unregister(String coreName,
                       CoreDescriptor cd,
                       boolean removeCoreFromZk)
                throws Exception

Throws:

Exception

createCollection

public void createCollection(String collection)
                      throws Exception

Throws:

Exception

getZkStateReader

public ZkStateReader getZkStateReader()

getCoreNodeName

public String getCoreNodeName(CoreDescriptor descriptor)

preRegister

public void preRegister(CoreDescriptor cd,
                        boolean publishState)

linkConfSet

public static void linkConfSet(SolrZkClient zkClient,
                               String collection,
                               String confSetName)
                        throws org.apache.zookeeper.KeeperException,
                               InterruptedException

Throws:

org.apache.zookeeper.KeeperException

InterruptedException

bootstrapConf

public static void bootstrapConf(SolrZkClient zkClient,
                                 CoreContainer cc,
                                 String solrHome)
                          throws IOException

If in SolrCloud mode, upload config sets for each SolrCore in solr.xml.

Throws:

IOException

getOverseerJobQueue

public ZkDistributedQueue getOverseerJobQueue()

getOverseerCollectionQueue

public OverseerTaskQueue getOverseerCollectionQueue()

getOverseerConfigSetQueue

public OverseerTaskQueue getOverseerConfigSetQueue()

getOverseerRunningMap

public DistributedMap getOverseerRunningMap()

getOverseerCompletedMap

public DistributedMap getOverseerCompletedMap()

getOverseerFailureMap

public DistributedMap getOverseerFailureMap()

claimAsyncId

public boolean claimAsyncId(String asyncId)
                     throws org.apache.zookeeper.KeeperException

When an operation needs to be performed in an asynchronous mode, the asyncId needs to be claimed by calling this method to make sure it's not duplicate (hasn't been claimed by other request). If this method returns true, the asyncId in the parameter has been reserved for the operation, meaning that no other thread/operation can claim it. If for whatever reason, the operation is not scheduled, the asuncId needs to be cleared using clearAsyncId(String). If this method returns false, no reservation has been made, and this asyncId can't be used, since it's being used by another operation (currently or in the past)

Parameters:

asyncId - A string representing the asyncId of an operation. Can't be null.

Returns:

True if the reservation succeeds. False if this ID is already in use.

Throws:

org.apache.zookeeper.KeeperException

clearAsyncId

public boolean clearAsyncId(String asyncId)
                     throws org.apache.zookeeper.KeeperException

Clears an asyncId previously claimed by calling claimAsyncId(String)

Parameters:

asyncId - A string representing the asyncId of an operation. Can't be null.

Returns:

True if the asyncId existed and was cleared. False if the asyncId didn't exist before.

Throws:

org.apache.zookeeper.KeeperException

getClientTimeout

public int getClientTimeout()

getOverseer

public Overseer getOverseer()

getOverseerElector

public LeaderElector getOverseerElector()

trimLeadingAndTrailingSlashes

public static String trimLeadingAndTrailingSlashes(String in)

Utility method for trimming and leading and/or trailing slashes from its input. May return the empty string. May return null if and only if the input is null.

rejoinOverseerElection

public void rejoinOverseerElection(String electionNode,
                                   boolean joinAtHead)

rejoinShardLeaderElection

public void rejoinShardLeaderElection(SolrParams params)

checkOverseerDesignate

public void checkOverseerDesignate()

getCoreContainer

public CoreContainer getCoreContainer()

ensureReplicaInLeaderInitiatedRecovery

@Deprecated
public boolean ensureReplicaInLeaderInitiatedRecovery(CoreContainer container,
                                                                  String collection,
                                                                  String shardId,
                                                                  ZkCoreNodeProps replicaCoreProps,
                                                                  CoreDescriptor leaderCd,
                                                                  boolean forcePublishState)
                                                           throws org.apache.zookeeper.KeeperException,
                                                                  InterruptedException

Deprecated.

When a leader receives a communication error when trying to send a request to a replica, it calls this method to ensure the replica enters recovery when connectivity is restored.

returns true if the node hosting the replica is still considered "live" by ZooKeeper; false means the node is not live either, so no point in trying to send recovery commands to it.

Throws:

org.apache.zookeeper.KeeperException

InterruptedException

isReplicaInRecoveryHandling

@Deprecated
public boolean isReplicaInRecoveryHandling(String replicaUrl)

Deprecated.

removeReplicaFromLeaderInitiatedRecoveryHandling

@Deprecated
public void removeReplicaFromLeaderInitiatedRecoveryHandling(String replicaUrl)

Deprecated.

getLeaderInitiatedRecoveryState

@Deprecated
public Replica.State getLeaderInitiatedRecoveryState(String collection,
                                                                 String shardId,
                                                                 String coreNodeName)

Deprecated.

getLeaderInitiatedRecoveryStateObject

@Deprecated
public Map<String,Object> getLeaderInitiatedRecoveryStateObject(String collection,
                                                                            String shardId,
                                                                            String coreNodeName)

Deprecated.

updateLeaderInitiatedRecoveryState

@Deprecated
public void updateLeaderInitiatedRecoveryState(String collection,
                                                           String shardId,
                                                           String coreNodeName,
                                                           Replica.State state,
                                                           CoreDescriptor leaderCd,
                                                           boolean retryOnConnLoss)

Deprecated.

getLeaderInitiatedRecoveryZnodePath

@Deprecated
public static String getLeaderInitiatedRecoveryZnodePath(String collection,
                                                                     String shardId)

Deprecated.

getLeaderInitiatedRecoveryZnodePath

@Deprecated
public static String getLeaderInitiatedRecoveryZnodePath(String collection,
                                                                     String shardId,
                                                                     String coreNodeName)

Deprecated.

throwErrorIfReplicaReplaced

public void throwErrorIfReplicaReplaced(CoreDescriptor desc)

addOnReconnectListener

public void addOnReconnectListener(OnReconnect listener)

Add a listener to be notified once there is a new session created after a ZooKeeper session expiration occurs; in most cases, listeners will be components that have watchers that need to be re-created.

removeOnReconnectListener

public void removeOnReconnectListener(OnReconnect listener)

Removed a previously registered OnReconnect listener, such as when a core is removed or reloaded.

persistConfigResourceToZooKeeper

public static int persistConfigResourceToZooKeeper(ZkSolrResourceLoader zkLoader,
                                                   int znodeVersion,
                                                   String resourceName,
                                                   byte[] content,
                                                   boolean createIfNotExists)

Persists a config file to ZooKeeper using optimistic concurrency.

Returns:

true on success

touchConfDir

public static void touchConfDir(ZkSolrResourceLoader zkLoader)

registerConfListenerForCore

public void registerConfListenerForCore(String confDir,
                                        SolrCore core,
                                        Runnable listener)

This will give a callback to the listener whenever a child is modified in the conf directory. It is the responsibility of the listener to check if the individual item of interest has been modified. When the last core which was interested in this conf directory is gone the listeners will be removed automatically.

getConfigDirListener

public OnReconnect getConfigDirListener()

checkIfCoreNodeNameAlreadyExists

public boolean checkIfCoreNodeNameAlreadyExists(CoreDescriptor dcore)

publishNodeAsDown

public void publishNodeAsDown(String nodeName)

Best effort to set DOWN state for all replicas on node.

Parameters:

nodeName - to operate on