fixed not to try to reconnect when session and connection timeout.

[youngwan657]

The application received the zookeeper event without any problems after network is reconnected.

e.g.
RetryPolicy retryPolicy = new ExponentialBackoffRetry(Config.BASE_SLEEP_TIME_MS, 3);
CuratorFrameworkFactory.newClient(connectString, 20000, 10000, retryPolicy);

But following example is not working after network is not available for about 30s.

e.g.
RetryPolicy retryPolicy = new ExponentialBackoffRetry(Config.BASE_SLEEP_TIME_MS, 3);
zkClient = CuratorFrameworkFactory.newClient(connectString, 10000, 10000, retryPolicy);

The application didn't receive any zookeeper event after this reset() method is invoked.

[Randgalt]

I don't understand this. This would break major Curator functionality.

[Randgalt]

Also, what Curator issue does this address?

[youngwan657]

I have a question.
How do I watch the znode again when reconnected after connection and session timeout?
I think the application didn't know RECONNECTED state.
Therefore the application didn't watch any change of the znode after that.

[log]
Connection attempt unsuccessful after 10205 (greater than max timeout of 10000). Resetting connection and trying again with a new connection.
reset
State change: RECONNECTED
There are no ConnectionStateListeners registered.

[Randgalt]

Watchers in ZooKeeper are one time only. Once the process() method is called, you need to reset the watch by calling whatever method you used originally. Alternatively, use one of the Curator recipes such as NodeCache.

[youngwan657]

But in this case, process() method wasn't called.
The application didn't detect any event.

That log is curator log not application log.

[Randgalt]

You can look at various Curator recipes. Generally, when RECONNECTED is received, Curator recipes reset whatever watchers are needed.