users@shoal.java.net

Re: Using the DSC

From: Joseph Fialli <joe.fialli_at_oracle.com>
Date: Wed, 09 Feb 2011 12:04:31 -0500

  Thomas,

In the log for A, I found the following events.

> 09.02.2011 09:42:14 com.sun.enterprise.ee.cms.impl.base.ViewWindowImpl
> newViewObserved
> INFO: GMS1016: Analyzing new membership snapshot received as part of
> event: ADD_EVENT for member: server9081 of group: default
> 09.02.2011 09:42:17
> com.sun.enterprise.ee.cms.impl.base.GroupCommunicationProviderImpl
> sendMessage
> WARNUNG: GroupCommunicationProvider.sendMessage(target=server9081):
> unable to send message: missing mapping from member identifier to
> network peerid
> 09.02.2011 09:42:17 com.sun.enterprise.ee.cms.impl.base.ViewWindowImpl
> addJoinNotificationSignal
What this indicates to me is that the send message that failed is quite
likely an attempt to synchronize the distributed state cache over TCP.
The DistributedStateCache synchronize is the first communication between
instances using TCP (other communications are UDP multicast).
A workaround is to delay setting DistributedStateCache values until the
group is up.

We can confirm the issue by enabling "ShoalLogger" to "FINER".
Add the following line to the java command starting the shoal gms client,
-Djava.util.logging.config.file=./logging.properties and add the line
"ShoalLogger FINER" to logging.properties file to enable ShoalLogger
logging at FINER level.
That log level will print out lots of info including distributed state
cache info.

Another workaround is to "get" values using
DistributedStateCache.getFromCacheForPattern(final String componentName,
final String memberToken).
This method not only looks in local cache but will double check if the
value has been set in the cluster by attempting to "pull" the value in.
The "pull" makes up for the failed "push" in the log file above.

-Joe

On 2/9/11 3:48 AM, Thomas Karolski wrote:
> Hi Shreedhar,
>
> thanks for the response, here is the output of both A and B attached
> to this mail.
>
> Kind regards,
> Thomas
>
> Hi Thomas
> Are the two instances being reported as seeing each other in the
> view in your logs?
>
> Could you share your log output?
>
> Thanks
> Shreedhar
>