EJB-RMI server hang using IIOP

Question

I have a java program that makes an EJB call using IIOP to an EJB hosted on a WebLogic server. The code is known to work in several environments, however when I deploy this to my production environment it hangs and appears to "lose" the reply from Weblogic.
Essentially, then the program makes an attempt to obtain the initial context, it sends a request to the Weblogic server, but then hangs.
Snooping the network shows that the request has reached weblogic and that weblogic has sent a sensible reply (which the client acknowledges).
From trussing the process I can see the program send the request (268 bytes of data written to a socket that matches with what I see in snoop), but the program then seems to be looping in ioctl and pollsys, I never see it read any data.
The last few lines of the stack trace show the process in an Object.wait (see code snippets).
The last call my code makes is to construct a javax/naming/InitialContext (see "context = new InitialContext(props);" in Code snippet)
Doing a "CTRL+\" does not show any deadlocked threads in the JVM.
Code Snippets:
The method in the code that it crashes in is:
* Helper method to obtain an initial context
private static void obtainInitialContext(boolean reload)
throws IOException
if (context == null || reload)
logger.info("Creating a new initial context as it either doesn't exist or needs reloading");
final String PROVIDER_URL_KEY = "props.orange.ProviderUrlKey";
final String CONTEXT_FACTORY_KEY = "props.orange.ContextFactoryKey";
try
String providerURL = ConfigurationManager.getConfigValue(PROVIDER_URL_KEY);
String contextFactory = ConfigurationManager.getConfigValue(CONTEXT_FACTORY_KEY);
if (providerURL == null)
logger.error("Could not find the property: " + PROVIDER_URL_KEY);
throw new IOException("Could not find the property: " + PROVIDER_URL_KEY);
if (contextFactory == null)
logger.error("Could not find the property: " + CONTEXT_FACTORY_KEY);
throw new IOException("Could not find the property: " + CONTEXT_FACTORY_KEY);
Hashtable props = new Hashtable();
props.put(Context.PROVIDER_URL, providerURL);
props.put(Context.INITIAL_CONTEXT_FACTORY, contextFactory);
context = new InitialContext(props);
catch (Exception e)
logger.error("Unable to obtain initial context: ", e);
throw new IOException("Unable to obtain initial context: " + e);
Stack trace:
ff2c1bbc lwp_cond_wait (3a8a0, 3a888, 0, 0)
feab4f98 __1cNObjectMonitorEwait6MxipnGThread__v_ (241d98, 94c8, 8c00, fefc4000, 39c08, 241db8) + 59c
feab61a8 __1cSObjectSynchronizerEwait6FnGHandle_xpnGThread__v_ (ffbfa758, 8784, 0, 39c08, 8400, fefc4000) + e8
feab5e58 JVM_MonitorWait (0, 3a204, 39c08, ff018ffc, 0, 1) + 31c
f900c280 * java/lang/Object.wait(J)V+0
f900c224 * java/lang/Object.wait(J)V+0
f9005764 * java/lang/Object.wait()V+2 (line 474)
f9005764 * com/sun/corba/se/impl/transport/CorbaResponseWaitingRoomImpl.waitForResponse(Lcom/sun/corba/se/pept/protocol/MessageMediator;)Lcom/sun/corba/se/pept/encoding/InputObject;+238 (line 140)

Answer 1

prabha kancharla <> writes:
What versions of each are you using?
andy
I have a java program that makes an EJB call using IIOP to an EJB hosted on a WebLogic server. The code is known to work in several environments, however when I deploy this to my production environment it hangs and appears to "lose" the reply from Weblogic.
Essentially, then the program makes an attempt to obtain the initial context, it sends a request to the Weblogic server, but then hangs.
Snooping the network shows that the request has reached weblogic and that weblogic has sent a sensible reply (which the client acknowledges).
From trussing the process I can see the program send the request (268 bytes of data written to a socket that matches with what I see in snoop), but the program then seems to be looping in ioctl and pollsys, I never see it read any data.
The last few lines of the stack trace show the process in an Object.wait (see code snippets).
The last call my code makes is to construct a javax/naming/InitialContext (see "context = new InitialContext(props);" in Code snippet)
Doing a "CTRL+\" does not show any deadlocked threads in the JVM.
Code Snippets:
The method in the code that it crashes in is:
* Helper method to obtain an initial context
private static void obtainInitialContext(boolean reload)
throws IOException
if (context == null || reload)
logger.info("Creating a new initial context as it either doesn't exist or needs reloading");
final String PROVIDER_URL_KEY = "props.orange.ProviderUrlKey";
final String CONTEXT_FACTORY_KEY = "props.orange.ContextFactoryKey";
try
String providerURL = ConfigurationManager.getConfigValue(PROVIDER_URL_KEY);
String contextFactory = ConfigurationManager.getConfigValue(CONTEXT_FACTORY_KEY);
if (providerURL == null)
logger.error("Could not find the property: " + PROVIDER_URL_KEY);
throw new IOException("Could not find the property: " + PROVIDER_URL_KEY);
if (contextFactory == null)
logger.error("Could not find the property: " + CONTEXT_FACTORY_KEY);
throw new IOException("Could not find the property: " + CONTEXT_FACTORY_KEY);
Hashtable props = new Hashtable();
props.put(Context.PROVIDER_URL, providerURL);
props.put(Context.INITIAL_CONTEXT_FACTORY, contextFactory);
context = new InitialContext(props);
catch (Exception e)
logger.error("Unable to obtain initial context: ", e);
throw new IOException("Unable to obtain initial context: " + e);
Stack trace:
ff2c1bbc lwp_cond_wait (3a8a0, 3a888, 0, 0)
feab4f98 __1cNObjectMonitorEwait6MxipnGThread__v_ (241d98, 94c8, 8c00, fefc4000, 39c08, 241db8) + 59c
feab61a8 __1cSObjectSynchronizerEwait6FnGHandle_xpnGThread__v_ (ffbfa758, 8784, 0, 39c08, 8400, fefc4000) + e8
feab5e58 JVM_MonitorWait (0, 3a204, 39c08, ff018ffc, 0, 1) + 31c
f900c280 * java/lang/Object.wait(J)V+0
f900c224 * java/lang/Object.wait(J)V+0
f9005764 * java/lang/Object.wait()V+2 (line 474)
f9005764 * com/sun/corba/se/impl/transport/CorbaResponseWaitingRoomImpl.waitForResponse(Lcom/sun/corba/se/pept/protocol/MessageMediator;)Lcom/sun/corba/se/pept/encoding/InputObject;+238 (line 140)