Class SlaveController

java.lang.Object
org.apache.derby.impl.store.replication.slave.SlaveController
All Implemented Interfaces:
ModuleControl, ModuleSupportable, SlaveFactory

public class SlaveController extends Object implements SlaveFactory, ModuleControl, ModuleSupportable

This is an implementation of the replication slave controller service. The service is booted when this instance of Derby will have the replication slave role for this database.

Note: The current version of the class is far from complete. Code to control the replication slave behavior will be added as more parts of the replication functionality is added to Derby.

See Also:
  • Field Details

    • DEFAULT_SOCKET_TIMEOUT

      private static final int DEFAULT_SOCKET_TIMEOUT
      See Also:
    • rawStoreFactory

      private RawStoreFactory rawStoreFactory
    • logToFile

      private LogToFile logToFile
    • receiver

      private ReplicationMessageReceive receiver
    • repLogger

      private ReplicationLogger repLogger
    • slaveAddr

      private SlaveAddress slaveAddr
    • dbname

      private String dbname
    • highestLogInstant

      private volatile long highestLogInstant
      The instant of the latest log record received from the master and processed so far. Used to check that master and slave log files are in synch
    • inReplicationSlaveMode

      private volatile boolean inReplicationSlaveMode
      Whether or not replication slave mode is still on. Will be set to false when slave replication is shut down. The value of this variable is checked after every timeout when trying to set up a connection to the master, and by the thread that applies log chunks received from the master.
    • startupSuccessful

      private volatile boolean startupSuccessful
      Whether or not this SlaveController has been successfully started, including setting up a connection with the master and starting the log receiver thread. The client connection that initiated slave replication mode on this database will not report that slave mode was successfully started (i.e., it will hang) until startupSuccessful has been set to true
    • logScan

      private ReplicationLogScan logScan
    • logReceiverThread

      private SlaveController.SlaveLogReceiverThread logReceiverThread
  • Constructor Details

    • SlaveController

      public SlaveController()
      Empty constructor required by Monitor.bootServiceModule
  • Method Details

    • boot

      public void boot(boolean create, Properties properties) throws StandardException
      Used by Monitor.bootServiceModule to start the service. It will set up basic variables
      Specified by:
      boot in interface ModuleControl
      Parameters:
      create - Currently ignored
      properties - Properties used to start the service in the correct mode
      Throws:
      StandardException - Standard Derby exception policy, thrown on error.
      See Also:
    • stop

      public void stop()
      Will tear down the replication slave service.
      Specified by:
      stop in interface ModuleControl
      See Also:
    • canSupport

      public boolean canSupport(Properties startParams)
      Used by Monitor.bootServiceModule to check if this class is usable for replication. To be usable, we require that slave replication mode is specified in startParams by checking that a property with key SlaveFactory.REPLICATION_MODE has the value SlaveFactory.SLAVE_MODE.
      Specified by:
      canSupport in interface ModuleSupportable
      Parameters:
      startParams - The properties used to start replication
      Returns:
      true if slave repliation is specified, meaning that this MasterController is a suitable implementation for the SlaveFactory service. False otherwise.
      See Also:
    • startSlave

      public void startSlave(RawStoreFactory rawStore, LogFactory logFac) throws StandardException
      Start slave replication. This method establishes a network connection with the associated replication master and starts a thread that applies operations received from the master (in the form of log records) to the local slave database.
      Specified by:
      startSlave in interface SlaveFactory
      Parameters:
      rawStore - The RawStoreFactory for the database
      logFac - The LogFactory ensuring recoverability for this database
      Throws:
      StandardException - Thrown if the slave could not be started.
    • stopSlave

      private void stopSlave() throws StandardException
      Will perform all work that is needed to stop replication
      Throws:
      StandardException
    • stopSlave

      public void stopSlave(boolean forcedStop) throws StandardException
      Description copied from interface: SlaveFactory
      Stop replication slave mode. Causes the database to abort the boot process, and should only be used when shutting down this database. If forcedStop is false, the method will fail with an exception if connected with the master. If forcedStop is true, the slave will be shut down even if connected to the master. A forcedStop value of true should only be used by system shutdown.
      Specified by:
      stopSlave in interface SlaveFactory
      Parameters:
      forcedStop - Determines whether or not an exception should be thrown when this method is called while the network connection to the master is up.
      Throws:
      StandardException - Thrown if slave is connected with master and forcedStop is false.
      See Also:
    • failover

      public void failover() throws StandardException
      Description copied from interface: SlaveFactory

      Used to turn this slave instance of the database into a normal instance that clients can connect to, assuming that the connection with the master is down. This is typically done in cases where a fatal error has happened on the master instance of the database, or when the master database is unreachable due to network problems.

      By calling failover, this slave instance of the database will be recovered so that all committed operations that have been received from the master are reflected here. On the other hand, operations from transactions where the commit log record has not been received from the master will not be reflected.

      Note that even though an operation has been executed (and even committed) on the master, it is not neccessarily reflected in the slave instance of the database. This depends on the replication strategy used by the MasterFactory.

      Specified by:
      failover in interface SlaveFactory
      Throws:
      StandardException - Thrown if slave is connected with master
      See Also:
    • doFailover

      private void doFailover()
      Performs failover on this database. May be called because a failover command has been received from the master, or because a client has requested a failover after the network connection with the master has been lost.
      See Also:
    • isStarted

      public boolean isStarted()
      Description copied from interface: SlaveFactory
      Check whether or not slave replication mode has been successfully started.
      Specified by:
      isStarted in interface SlaveFactory
      Returns:
      true if slave replication mode has started successfully, false if slave mode startup is not yet confirmed
      See Also:
    • setupConnection

      private boolean setupConnection() throws StandardException
      Establish a connection with the replication master. Listens for a connection on the slavehost/port for DEFAULT_SOCKET_TIMEOUT milliseconds.
      Returns:
      true if a connection has been set up with the master, false if the connection attempt timed out.
      Throws:
      StandardException - if an unexpected exception occured that prevented a connection with the master.
    • handleDisconnect

      private void handleDisconnect(Exception e)
      Write the reason for the lost connection to the log (derby.log) and reconnect with the master. Once the network is up and running, a new LogReceiverThread is started. The method returns without doing anything if inReplicationSlaveMode=false, which means that stopSlave() has been called by another thread.
      Parameters:
      e - The reason the connection to the master was lost
    • isConnectedToMaster

      private boolean isConnectedToMaster()
      Check if the repliation network connection to the master is working
      Returns:
      true if the network connection is working, false otherwise
    • startLogReceiverThread

      private void startLogReceiverThread()
      Starts the LogReceiverThread that will listen for chunks of log records from the master and apply the log records to the local log file.
    • handleFatalException

      private void handleFatalException(Exception e)
      Handles fatal errors for slave replication functionality. These are errors that requires us to stop replication. Calling this method has the following effects: 1) Debug messages are written to the log file (usually derby.log) if ReplicationLogger#LOG_REPLICATION_MESSAGES is true. 2) If the network connection is up, the master is notified of the problem. 3) All slave replication functionality is stopped, and the database is then shut down without being booted. The method will return without doing anything if inReplicationSlaveMode=false, meaning that stopSlave has been called.
      Parameters:
      e - The fatal exception that is the reason for calling this method
    • teardownNetwork

      private void teardownNetwork()
    • getHostName

      private String getHostName()
      Used to return the host name of the slave.
      Returns:
      a String containing the host name of the slave.
    • getPortNumber

      private int getPortNumber()
      Used to return the port number of the slave.
      Returns:
      an Integer that represents the port number of the slave.