org.apache.ignite.internal.cluster

Class IgniteClusterImpl

java.lang.Object

org.apache.ignite.internal.cluster.ClusterGroupAdapter

org.apache.ignite.internal.cluster.IgniteClusterImpl

All Implemented Interfaces:

Externalizable, Serializable, ClusterGroup, IgniteCluster, ClusterGroupEx, IgniteClusterEx, IgniteAsyncSupport

public class IgniteClusterImpl
extends ClusterGroupAdapter
implements IgniteClusterEx, Externalizable

See Also:

Serialized Form

Field Summary

Fields inherited from class org.apache.ignite.internal.cluster.ClusterGroupAdapter

ctx, p, subjId

Constructor Summary

Constructors

Constructor and Description
IgniteClusterImpl() Required by Externalizable.
IgniteClusterImpl(GridKernalContext ctx)

Method Summary

Methods

Modifier and Type	Method and Description
void	clearNodeMap() Clears node local map.
ClusterGroup	forLocal() Gets a cluster group consisting from the local node.
<R> IgniteFuture<R>	future() Gets and resets future for previous asynchronous operation.
boolean	isAsync()
ClusterNode	localNode() Gets local grid node.
<K> Map<ClusterNode,Collection<K>>	mapKeysToNodes(String cacheName, Collection<? extends K> keys) This method provides ability to detect which cache keys are mapped to which nodes on cache instance with given name.
<K> ClusterNode	mapKeyToNode(String cacheName, K key) This method provides ability to detect which cache keys are mapped to which nodes on cache instance with given name.
<K,V> ConcurrentMap<K,V>	nodeLocalMap() Gets node-local storage instance.
boolean	pingNode(UUID nodeId) Pings a remote node.
void	readExternal(ObjectInput in)
protected Object	readResolve() Reconstructs object on unmarshalling.
void	resetMetrics() Resets local I/O, job, and task execution metrics.
void	restartNodes() Restarts nodes satisfying optional set of predicates.
void	restartNodes(Collection<UUID> ids) Restarts nodes defined by provided IDs.
Collection<ClusterStartNodeResult>	startNodes(Collection<Map<String,Object>> hosts, Map<String,Object> dflts, boolean restart, int timeout, int maxConn) Starts one or more nodes on remote host(s).
Collection<ClusterStartNodeResult>	startNodes(File file, boolean restart, int timeout, int maxConn) Starts one or more nodes on remote host(s).
void	stopNodes() Stops nodes satisfying optional set of predicates.
void	stopNodes(Collection<UUID> ids) Stops nodes defined by provided IDs.
Collection<ClusterNode>	topology(long topVer) Gets a topology by version.
long	topologyVersion() Gets current topology version.
String	toString()
IgniteCluster	withAsync() Gets instance of this component with asynchronous mode enabled.
void	writeExternal(ObjectOutput out)

Methods inherited from class org.apache.ignite.internal.cluster.ClusterGroupAdapter

compute, events, executorService, forAttribute, forCacheNodes, forCacheNodes, forClientNodes, forClients, forDaemons, forDataNodes, forHost, forNode, forNodeId, forNodeIds, forNodes, forOldest, forOthers, forOthers, forPredicate, forRandom, forRemotes, forServers, forSubjectId, forYoungest, guard, ignite, message, metrics, node, node, nodes, predicate, services, setKernalContext, unguard

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.apache.ignite.internal.cluster.ClusterGroupEx

forCacheNodes, forSubjectId

Methods inherited from interface org.apache.ignite.cluster.ClusterGroup

forAttribute, forCacheNodes, forClientNodes, forClients, forDaemons, forDataNodes, forHost, forNode, forNodeId, forNodeIds, forNodes, forOldest, forOthers, forOthers, forPredicate, forRandom, forRemotes, forServers, forYoungest, ignite, metrics, node, node, nodes, predicate

Constructor Detail

IgniteClusterImpl

public IgniteClusterImpl()

Required by Externalizable.

IgniteClusterImpl

public IgniteClusterImpl(GridKernalContext ctx)

Parameters:

ctx - Kernal context.

Method Detail

forLocal

public ClusterGroup forLocal()

Gets a cluster group consisting from the local node.

Specified by:

forLocal in interface IgniteCluster

Returns:

Cluster group consisting from the local node.

localNode

public ClusterNode localNode()

Gets local grid node.

Specified by:

localNode in interface IgniteCluster

Returns:

Local grid node.

nodeLocalMap

public <K,V> ConcurrentMap<K,V> nodeLocalMap()

Gets node-local storage instance.

Node-local values are similar to thread locals in a way that these values are not distributed and kept only on local node (similar like thread local values are attached to the current thread only). Node-local values are used primarily by closures executed from the remote nodes to keep intermediate state on the local node between executions.

There's only one instance of node local storage per local node. Node local storage is based on ConcurrentMap and is safe for multi-threaded access.

Specified by:

nodeLocalMap in interface IgniteCluster

Returns:

Node local storage instance for the local node.

pingNode

public boolean pingNode(UUID nodeId)

Pings a remote node.

Discovery SPIs usually have some latency in discovering failed nodes. Hence, communication to remote nodes may fail at times if an attempt was made to establish communication with a failed node. This method can be used to check if communication has failed due to node failure or due to some other reason.

Specified by:

pingNode in interface IgniteCluster

Parameters:

nodeId - ID of a node to ping.

Returns:

true if node for a given ID is alive, false otherwise.

See Also:

DiscoverySpi

topologyVersion

public long topologyVersion()

Gets current topology version. In case of TCP discovery (see TcpDiscoverySpi) topology versions are sequential - they start from '1' and get incremented every time whenever a node joins or leaves. For other discovery SPIs topology versions may not be (and likely are not) sequential.

Specified by:

topologyVersion in interface IgniteCluster

Returns:

Current topology version.

topology

public Collection<ClusterNode> topology(long topVer)
                                 throws UnsupportedOperationException

Gets a topology by version. Returns null if topology history storage doesn't contain specified topology version (history currently keeps last 1000 snapshots).

Specified by:

topology in interface IgniteCluster

Parameters:

topVer - Topology version.

Returns:

Collection of grid nodes which represented by specified topology version, if it is present in history storage, null otherwise.

Throws:

UnsupportedOperationException - If underlying SPI implementation does not support topology history. Currently only TcpDiscoverySpi supports topology history.

mapKeysToNodes

public <K> Map<ClusterNode,Collection<K>> mapKeysToNodes(@Nullable
                                                String cacheName,
                                                @Nullable
                                                Collection<? extends K> keys)
                                              throws IgniteException

This method provides ability to detect which cache keys are mapped to which nodes on cache instance with given name. Use it to determine which nodes are storing which keys prior to sending jobs that access these keys.

This method works as following:

• For local caches it returns only local node mapped to all keys.
• For fully replicated caches, AffinityFunction is used to determine which keys are mapped to which groups of nodes.
• For partitioned caches, the returned map represents node-to-key affinity.

Specified by:

mapKeysToNodes in interface IgniteCluster

Parameters:

cacheName - Cache name, if null, then default cache instance is used.

keys - Cache keys to map to nodes.

Returns:

Map of nodes to cache keys or empty map if there are no alive nodes for this cache.

Throws:

IgniteException - If failed to map cache keys.

mapKeyToNode

public <K> ClusterNode mapKeyToNode(@Nullable
                           String cacheName,
                           K key)
                         throws IgniteException

This method provides ability to detect which cache keys are mapped to which nodes on cache instance with given name. Use it to determine which nodes are storing which keys prior to sending jobs that access these keys.

This method works as following:

• For local caches it returns only local node ID.
• For fully replicated caches first node ID returned by AffinityFunction is returned.
• For partitioned caches, the returned node ID is the primary node for the key.

Specified by:

mapKeyToNode in interface IgniteCluster

Parameters:

cacheName - Cache name, if null, then default cache instance is used.

key - Cache key to map to a node.

Returns:

Primary node for the key or null if cache with given name is not present in the grid.

Throws:

IgniteException - If failed to map key.

startNodes

public Collection<ClusterStartNodeResult> startNodes(File file,
                                            boolean restart,
                                            int timeout,
                                            int maxConn)
                                              throws IgniteException

Starts one or more nodes on remote host(s).

This method takes INI file which defines all startup parameters. It can contain one or more sections, each for a host or for range of hosts (note that they must have different names) and a special 'defaults' section with default values. They are applied to undefined parameters in host's sections.

Returned result is collection of tuples. Each tuple corresponds to one node start attempt and contains hostname, success flag and error message if attempt was not successful. Note that successful attempt doesn't mean that node was actually started and joined topology. For large topologies (> 100s nodes) it can take over 10 minutes for all nodes to start. See individual node logs for details.

Supports asynchronous execution (see IgniteAsyncSupport).

Specified by:

startNodes in interface IgniteCluster

Parameters:

file - Configuration file.

restart - Whether to stop existing nodes. If true, all existing nodes on the host will be stopped before starting new ones. If false, nodes will be started only if there are less nodes on the host than expected.

timeout - Connection timeout.

maxConn - Number of parallel SSH connections to one host.

Returns:

Collection of start node results, each containing host name, result (success or failure) and error message (if any).

Throws:

IgniteException - In case of error.

startNodes

public Collection<ClusterStartNodeResult> startNodes(Collection<Map<String,Object>> hosts,
                                            @Nullable
                                            Map<String,Object> dflts,
                                            boolean restart,
                                            int timeout,
                                            int maxConn)
                                              throws IgniteException

Starts one or more nodes on remote host(s).

Each map in hosts collection defines startup parameters for one host or for a range of hosts. The following parameters are supported:

Name	Type	Description
host	String	Hostname (required). Can define several hosts if their IPs are sequential. E.g., 10.0.0.1~5 defines range of five IP addresses. Other parameters are applied to all hosts equally.
port	Integer	Port number (default is 22).
uname	String	Username (if not defined, current local username will be used).
passwd	String	Password (if not defined, private key file must be defined).
key	File	Private key file (if not defined, password must be defined).
nodes	Integer	Expected number of nodes on the host. If some nodes are started already, then only remaining nodes will be started. If current count of nodes is equal to this number, and restart flag is false, then nothing will happen.
igniteHome	String	Path to Ignite installation folder. If not defined, IGNITE_HOME environment variable must be set on remote hosts.
cfg	String	Path to configuration file (relative to igniteHome).
script	String	Custom startup script file name and path (relative to igniteHome). You can also specify a space-separated list of parameters in the same string (for example: "bin/my-custom-script.sh -v").

dflts map defines default values. They are applied to undefined parameters in hosts collection.

Returned result is collection of tuples. Each tuple corresponds to one node start attempt and contains hostname, success flag and error message if attempt was not successful. Note that successful attempt doesn't mean that node was actually started and joined topology. For large topologies (> 100s nodes) it can take over 10 minutes for all nodes to start. See individual node logs for details.

Supports asynchronous execution (see IgniteAsyncSupport).

Specified by:

startNodes in interface IgniteCluster

Parameters:

hosts - Startup parameters.

dflts - Default values.

restart - Whether to stop existing nodes. If true, all existing nodes on the host will be stopped before starting new ones. If false, nodes will be started only if there are less nodes on the host than expected.

timeout - Connection timeout in milliseconds.

maxConn - Number of parallel SSH connections to one host.

Returns:

Collection of start node results, each containing host name, result (success or failure) and error message (if any).

Throws:

IgniteException - In case of error.

stopNodes

public void stopNodes()
               throws IgniteException

Stops nodes satisfying optional set of predicates.

NOTE: System.exit(Ignition.KILL_EXIT_CODE) will be executed on each stopping node. If you have other applications running in the same JVM along with Ignition, those applications will be stopped as well.

Specified by:

stopNodes in interface IgniteCluster

Throws:

IgniteException - In case of error.

stopNodes

public void stopNodes(Collection<UUID> ids)
               throws IgniteException

Stops nodes defined by provided IDs.

NOTE: System.exit(Ignition.KILL_EXIT_CODE) will be executed on each stopping node. If you have other applications running in the same JVM along with Ignition, those applications will be stopped as well.

Specified by:

stopNodes in interface IgniteCluster

Parameters:

ids - IDs defining nodes to stop.

Throws:

IgniteException - In case of error.

restartNodes

public void restartNodes()
                  throws IgniteException

Restarts nodes satisfying optional set of predicates.

NOTE: this command only works for grid nodes started with Ignition ignite.sh or ignite.bat scripts.

Specified by:

restartNodes in interface IgniteCluster

Throws:

IgniteException - In case of error.

restartNodes

public void restartNodes(Collection<UUID> ids)
                  throws IgniteException

Restarts nodes defined by provided IDs.

NOTE: this command only works for grid nodes started with Ignition ignite.sh or ignite.bat scripts.

Specified by:

restartNodes in interface IgniteCluster

Parameters:

ids - IDs defining nodes to restart.

Throws:

IgniteException - In case of error.

resetMetrics

public void resetMetrics()

Resets local I/O, job, and task execution metrics.

Specified by:

resetMetrics in interface IgniteCluster

withAsync

public IgniteCluster withAsync()

Gets instance of this component with asynchronous mode enabled.

Specified by:

withAsync in interface IgniteCluster

Specified by:

withAsync in interface IgniteAsyncSupport

Returns:

Instance of this component with asynchronous mode enabled.

isAsync

public boolean isAsync()

Specified by:

isAsync in interface IgniteAsyncSupport

Returns:

True if asynchronous mode is enabled.

future

public <R> IgniteFuture<R> future()

Gets and resets future for previous asynchronous operation.

Specified by:

future in interface IgniteAsyncSupport

Returns:

Future for previous asynchronous operation.

clearNodeMap

public void clearNodeMap()

Clears node local map.

readExternal

public void readExternal(ObjectInput in)
                  throws IOException,
                         ClassNotFoundException

Specified by:

readExternal in interface Externalizable

Overrides:

readExternal in class ClusterGroupAdapter

Throws:

IOException

ClassNotFoundException

writeExternal

public void writeExternal(ObjectOutput out)
                   throws IOException

Specified by:

writeExternal in interface Externalizable

Overrides:

writeExternal in class ClusterGroupAdapter

Throws:

IOException

readResolve

protected Object readResolve()
                      throws ObjectStreamException

Reconstructs object on unmarshalling.

Overrides:

readResolve in class ClusterGroupAdapter

Returns:

Reconstructed object.

Throws:

ObjectStreamException - Thrown in case of unmarshalling error.

toString

public String toString()

Overrides:

toString in class Object