[BOOKKEEPER-63] Hedwig PubSubServer must wait for its Zookeeper client to be connected upon startup

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 4.0.0
• Component/s: hedwig-server
• Assignee: Ivan Kelly
• Reporter: Matthieu Morel
• Created: 31/Aug/11 11:29
• Updated: 07/Dec/11 15:56
• Resolved: 05/Sep/11 10:33

Description

When a PubSubServer is instantiated in non-standalone mode, it creates a ZkTopicManager which takes a Zookeeper client as an argument.
Unfortunately, this Zookeeper client may not be connected yet (not in CONNECTED state yet), and when this is the case, creation of ZkTopicManager fails, leading to failure of the PubSubServer startup.

Typical error (adapted, line numbers take into account commented patching code):
jjava.io.IOException: org.apache.hedwig.exceptions.PubSubException$ServiceDownException: org.apache.zookeeper.KeeperException$ConnectionLossException: KeeperErrorCode = ConnectionLoss for /hedwig/standalone/hosts/x.x.x.x:4080:9876
at org.apache.hedwig.server.netty.PubSubServer.instantiateTopicManager(PubSubServer.java:170)
at org.apache.hedwig.server.netty.PubSubServer$3.run(PubSubServer.java:294)
at java.lang.Thread.run(Thread.java:680)
Caused by: org.apache.hedwig.exceptions.PubSubException$ServiceDownException: org.apache.zookeeper.KeeperException$ConnectionLossException: KeeperErrorCode = ConnectionLoss for /hedwig/standalone/hosts/x.x.x.x:4080:9876
at org.apache.hedwig.server.topics.ZkTopicManager$4.safeProcessResult(ZkTopicManager.java:146)
etc...

This is particularly problematic for running tests that require to pass a config to the PubSubServer.