Disk Full Error Message :

WARN  Arjuna.logging.arjLoggerI18N-[Com.arjuna.ats.arjuna.recovery.RecoverAtomicAction_4|com.arjuna.ats.arjuna.recovery.RecoverAtomicAction_4] - RecoverAtomicAction: transaction not activated, unable to replay phase 2 commit

The above WARNING appears when the writes of transactional object state to the persistent store have failed because the disk is full. This will create the corrupt records in the store, probably zero length files. The recovery process tries to retrieve the transactional object state so that it can recreate the relevant object in memory. This transaction object may be a transaction, a participant record or some other type of object. Since the records are corrupt this is going to keep failing and the recovery manager will keep retrying and failing.

If this occurs in a DEVELOPMENT server then user should stop their app server, delete any files found in the object store and then restart. Alternatively the users can just delete the whole object store and it will be recreated.

If this error appears in the PRODUCTION server then the user need to be much more careful.  Alongside the invalid records the store may also contain valid records containing information about in-flight transactions and in-doubt transactions. If that is the case then they need to deal with these before bringing the app server down and cleaning out the object store. This may involve stopping the AS from accepting new requests (so that there are no in-flight transactions when it shuts down) and resolving any in-doubt transactions -- either manually or by waiting for the transaction manager to resolve them. Depending upon what happens this may also involve dealing with heuristic outcomes.