The 3rd node will fail to join the cluster, if peerClassLoading is enable on CQ situation.

Wed, 03 Aug 2016 00:52:06 -0700 

Hi all,


I have met a problem when using continuous query with remote filter if p2p 
class loading is enabled. I have pasted the testing code in gist 
https://gist.github.com/jackeylu/edf600b66273d17db0438065ae362934 , and you can 
reproduce it with the follow operations,


1. start an Ignite server with class IgniteServer
2. start the CQ client with class CQListenerClient, it will create a continuous 
query cursor for the CacheEntryEvent. In my case, I have a remote filter named 
EventPrinter which has its externalization methods.
3. start the third node with class IgniteSimpleClient, and exception occurs.


[15:19:02,567][SEVERE][tcp-client-disco-msg-worker-#4%null%][TcpDiscoverySpi] 
Runtime error caught during grid runnable execution: IgniteSpiThread 
[name=tcp-client-disco-msg-worker-#4%null%]
java.lang.NullPointerException
        at 
org.apache.ignite.internal.processors.cache.query.continuous.CacheContinuousQueryHandlerV2.getEventFilter(CacheContinuousQueryHandlerV2.java:111)
        at 
org.apache.ignite.internal.processors.cache.query.continuous.CacheContinuousQueryHandler.register(CacheContinuousQueryHandler.java:319)
        at 
org.apache.ignite.internal.processors.continuous.GridContinuousProcessor.registerHandler(GridContinuousProcessor.java:1101)
        at 
org.apache.ignite.internal.processors.continuous.GridContinuousProcessor.onDiscoveryDataReceived(GridContinuousProcessor.java:470)
        at 
org.apache.ignite.internal.managers.discovery.GridDiscoveryManager$5.onExchange(GridDiscoveryManager.java:668)
        at 
org.apache.ignite.spi.discovery.tcp.TcpDiscoverySpi.onExchange(TcpDiscoverySpi.java:1723)
        at 
org.apache.ignite.spi.discovery.tcp.ClientImpl$MessageWorker.processNodeAddFinishedMessage(ClientImpl.java:1670)
        at 
org.apache.ignite.spi.discovery.tcp.ClientImpl$MessageWorker.processDiscoveryMessage(ClientImpl.java:1553)
        at 
org.apache.ignite.spi.discovery.tcp.ClientImpl$MessageWorker.body(ClientImpl.java:1478)
        at org.apache.ignite.spi.IgniteSpiThread.run(IgniteSpiThread.java:62)



my ignite version is ver. 1.6.0#20160518-sha1:0b22c45b. And log files for the 3 
nodes are
1. the first node, ignite-2e8beaca.0.log
2. the second node, ignite-f12fd69b.0.log
3. the third node, ignite-2c73bf53.0.log


I think, it is may be a bug of p2p class loading, when p2p is enabled, the 
remote filter factory will be deployed by a DeployableObject wrapper, but when 
the 3rd simple and normal node join the cluster, it does not get the 
remoteFilterFactory, it get the remoteFilterFactoryDep from readExternal method 
which cause the remoteFilterFactory is null, and cause the NULLPointerException.


I think, if the third node can get the remoteFilterFactory with the method 
p2pUnmarshall in 
org.apache.ignite.internal.processors.cache.query.continuous.CacheContinuousQueryHandlerV2,
 then the problem can be solved, but currently it is not.


If I turn off the peer class loading, the remoteFilterFactory can be serialized 
and deserialized with the writeExternal/readExternal method, and everything 
will be OK. But, I can't do it like this, because in my case, the remote filter 
is coming from some third party jars,  they are not implemented JDK 
java.io.Serializableor java.io.Externalizable interfaces.


Any advice are welcomed, thank you so much.


Regards,


Lin.

Attachment: ignite-2c73bf53.0.log
Description: Binary data

Attachment: ignite-2e8beaca.0.log
Description: Binary data

Attachment: ignite-f12fd69b.0.log
Description: Binary data

Reply via email to