Atomikos Forum |
|
Case is as follows: A box runs out of disk. TE tries to write the epoch file, which ends up empty. Someone tries to restart with the 0-byte epoch file in place, which causes the following stack trace:
java.lang.RuntimeException at com.atomikos.util.UniqueIdMgr.<init>(UniqueIdMgr.java:112) at com.atomikos.util.UniqueIdMgr.<init>(UniqueIdMgr.java:88) at com.atomikos.icatch.standalone.StandAloneTransactionManager.<init>(StandAloneTransactionManager.java:82) (...) The underlying cause is trying to do a readLong on the empty epoch file. The original error message is mangled by catching and rethrowing the exception. I would have liked the error message to state that the epoch file was corrupted, and that the system should be stopped, the file removed after which the system can be restarted. |