On 06/ 9/10 06:38 PM, Michael Closson wrote:Yes, and there's also a risk that parts of transactions are lost and
> Perhaps derby wrote an incomplete log record, and now cannot recover?
> Also, if I delete the log control file (log.ctrl) then I can connect
> OK. But if I do this, is there risk that some transactions will be
> lost? (I think so).
that uncommitted updates are persisted. It might work as a last resort
to recover data from the database, but the integrity of the data cannot
be guaranteed. If the log is corrupt, recovery from the last good backup
is the only safe way to get the database back into a consistent state.
DERBY-3896 is a problem that only affects replication, so I think that's
> Anyways, I looked for existing bugs similar to this one and the
> closest thing I can find is DERBY-3896
> (https://issues.apache.org/jira/browse/DERBY-3896), which is still open.
> Shall I log this as a new bug?
a different bug, even if the same assert is triggered. A new bug report
would be best.
Just attaching what you have of logs and stack traces first is fine. But
> Lastly, I have the database that causes this error, but I prefer to
> not post it to the bug log or mailing list. But if someone emails me
> directly, I will supply it.
please keep a copy of the database in case someone needs more info when
they start investigating the issue.