Package com.mapr.baseutils.zookeeper

Class ZKUtils

java.lang.Object

com.mapr.baseutils.zookeeper.ZKUtils

public class ZKUtils
extends java.lang.Object

Class to take care of zookeeper handling utilities with "connection loss" and "session expiration" handling to do: getData check for znode existence getChildren

Field Summary

Fields

Modifier and Type	Field	Description
static java.lang.String	SERVICE_MASTER_NODE
static int	TIMEOUT
static int	TIMEOUT_SERVER
static int	ZK_RETRIALS

Constructor Summary

Constructors

Constructor	Description
ZKUtils()

Method Summary

Modifier and Type	Method	Description
static org.apache.zookeeper.data.Stat	checkZKNodeForExistence(org.apache.zookeeper.ZooKeeper s_zk, java.lang.String nodeName, org.apache.zookeeper.Watcher watcher, int retrials)	Helper method to check for node existence in ZK with retrial logic in case of connection loss
static byte[]	getData(org.apache.zookeeper.ZooKeeper s_zk, java.lang.String nodeName, org.apache.zookeeper.Watcher watcher, org.apache.zookeeper.data.Stat stat, int retrials)	Helper method to get data from ZK with retrial logic in case of connection loss
static java.util.List<java.lang.String>	getZkNodeChildren(org.apache.zookeeper.ZooKeeper s_zk, java.lang.String nodeName, org.apache.zookeeper.Watcher watcher, int retrials)	Helper method to get znode children from ZK with retrial logic in case of connection loss
static void	resetZookeeper(java.lang.String zkAddresses, org.apache.zookeeper.ZooKeeper s_zk, org.apache.zookeeper.Watcher watcher)	Method to reset ZK - in case of KeeperException.SessionExpiredException that exception should actually remove znode from ZK - at least for nodes, as they are going to be EPHEMERAL_SEQUENTIAL, so we need to make sure that even new "master" is in place because of this (worst case master got SessionExpiredException) we do not restart service on any other node if it is running on "this one" Using writelock to prevent "reading" threads to get data that is in inconsistent state - during reset

Methods inherited from class java.lang.Object

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

Field Detail

ZK_RETRIALS

public static final int ZK_RETRIALS

See Also:

Constant Field Values

TIMEOUT

public static final int TIMEOUT

See Also:

Constant Field Values

TIMEOUT_SERVER

public static final int TIMEOUT_SERVER

See Also:

Constant Field Values

SERVICE_MASTER_NODE

public static final java.lang.String SERVICE_MASTER_NODE

See Also:

Constant Field Values

Constructor Detail

ZKUtils

public ZKUtils()

Method Detail

getData

public static byte[] getData(org.apache.zookeeper.ZooKeeper s_zk,
                             java.lang.String nodeName,
                             org.apache.zookeeper.Watcher watcher,
                             org.apache.zookeeper.data.Stat stat,
                             int retrials)
                      throws org.apache.zookeeper.KeeperException,
                             java.lang.InterruptedException,
                             ZKClosedException

Helper method to get data from ZK with retrial logic in case of connection loss

Parameters:

s_zk -

nodeName -

watcher -

stat -

retrials -

Returns:

Throws:

org.apache.zookeeper.KeeperException

java.lang.InterruptedException

ZKClosedException

checkZKNodeForExistence

public static org.apache.zookeeper.data.Stat checkZKNodeForExistence(org.apache.zookeeper.ZooKeeper s_zk,
                                                                     java.lang.String nodeName,
                                                                     org.apache.zookeeper.Watcher watcher,
                                                                     int retrials)
                                                              throws org.apache.zookeeper.KeeperException,
                                                                     java.lang.InterruptedException,
                                                                     ZKClosedException

Helper method to check for node existence in ZK with retrial logic in case of connection loss

Parameters:

s_zk -

nodeName -

watcher -

retrials -

Returns:

Throws:

org.apache.zookeeper.KeeperException

java.lang.InterruptedException

ZKClosedException

getZkNodeChildren

public static java.util.List<java.lang.String> getZkNodeChildren(org.apache.zookeeper.ZooKeeper s_zk,
                                                                 java.lang.String nodeName,
                                                                 org.apache.zookeeper.Watcher watcher,
                                                                 int retrials)
                                                          throws org.apache.zookeeper.KeeperException,
                                                                 java.lang.InterruptedException,
                                                                 ZKClosedException

Helper method to get znode children from ZK with retrial logic in case of connection loss

Parameters:

s_zk -

nodeName -

watcher -

retrials -

Returns:

Throws:

org.apache.zookeeper.KeeperException

java.lang.InterruptedException

ZKClosedException

resetZookeeper

public static void resetZookeeper(java.lang.String zkAddresses,
                                  org.apache.zookeeper.ZooKeeper s_zk,
                                  org.apache.zookeeper.Watcher watcher)
                           throws java.io.IOException

Method to reset ZK - in case of KeeperException.SessionExpiredException that exception should actually remove znode from ZK - at least for nodes, as they are going to be EPHEMERAL_SEQUENTIAL, so we need to make sure that even new "master" is in place because of this (worst case master got SessionExpiredException) we do not restart service on any other node if it is running on "this one" Using writelock to prevent "reading" threads to get data that is in inconsistent state - during reset

Parameters:

s_zk -

watcher -

Throws:

java.io.IOException