You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@zookeeper.apache.org by "Hongchao Deng (JIRA)" <ji...@apache.org> on 2014/09/15 23:01:34 UTC
[jira] [Updated] (ZOOKEEPER-2037) ZooKeeper methods to wait on
client connection (re)establishment.
[ https://issues.apache.org/jira/browse/ZOOKEEPER-2037?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Hongchao Deng updated ZOOKEEPER-2037:
-------------------------------------
Attachment: ZOOKEEPER-2037.patch
> ZooKeeper methods to wait on client connection (re)establishment.
> -----------------------------------------------------------------
>
> Key: ZOOKEEPER-2037
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2037
> Project: ZooKeeper
> Issue Type: Sub-task
> Components: java client
> Reporter: Hongchao Deng
> Assignee: Hongchao Deng
> Attachments: ZOOKEEPER-2037.patch
>
>
> When a ZooKeeper object is created and returned, it is not guaranteed that connection to a server is established.
> Usually, a wait/signal pattern is used for the ZK watcher
> {code}
> latch = new CountDownLatch(1)
> zk = new ZooKeeper(..., new Watcher() {
> override void process(WatchedEvent event) {
> if (event.type = SyncConnected) {
> latch.countDown()
> }
> }
> },...)
> latch.wait();
> // connection has been established. do something with zk.
> {code}
> There are two disadvantages:
> 1. The latch object isn't being garbage-collected. Because the watcher keeps monitoring all kinds of events.
> 2. With the introduction of dynamic reconfig, client move to other servers on needed and this latch method doesn't work so well.
> Here I propose to add (both sync and async) wait methods to act as latch for connection establish such that it becomes much easier to manage and work around:
> {code}
> zk = new ZooKeeper(...)
> zk.waitUntilConnected()
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)