Restarting Coherence

Question

Dear Coherents,
We have several application-level tasks running inside a single JVM. Each task is run in its own thread but tasks are run serially, so that only one task is running at any point in time (actually managed by DataSynapse). Each of the tasks may use a different Coherence configuration and therefore begins by calling:
1: CacheFactory.shutdown();
2: System.setProperty("tangosol.coherence.override", /* config file for current function */);
3: System.setProperty("tangosol.coherence.cacheconfig", /* config file for current function */);
4: CacheFactory.ensureCluster();
These calls are made in a single thread for each task, but occassionally Coherence throws an assertion error on line 4 with the message "run() invoked on a different thread".
I expect this is because CacheFactory.shutdown() returns before daemon Coherence threads from the previous task are completely stopped, and so for a moment there are two sets of Coherence threads running simultaneously; or at least previous static data has not been fully cleared.
Putting a 10 second sleep statement between lines 1 and 2 appears to resolve the issue, but any timeout-based synchronization is essentially bad practice in my book.
I would like to know firstly whether this assertion error signifies a problem, although the cluster appears to operate fine aftwards. Secondly, what is your recommendation to guarantee that this does not occur?
On a bigger note, is there any way to get visibility on:
- whether the JVM is currently a member of Coherence (apart from 'ensuring' that it is);
- what Coherence daemon threads are running;
- how long Coherence daemon threads have been running.
Many thanks,
James

Answer 1

James,
Please send the full logs to our support team at Oracle Metalink: (https://metalink.oracle.com)
Regards,
Gene