Hope your doing fine. Bhargava and I have come up with one more change along with the changes for the fix. Can you please let us know if this design is feasible and acceptable to the log4j flume appender community :
To skip sending events when the environment (Berkley DB or NFS) is down. To create a new database environment once the environment is back. We will loose the events in this case when the environment is down. — I have made code changes for this issue. I am finding it hard to recreate the issue. I am currently working on the unit test cases.
To switch to synchronous transfer of events when there is a problem with the database/NFS. This change is to prevent the loss of events. Once the environment is back, the code switches to Asynchronous mode. Please let us know your opinion on this.