Good bug report is a key for fast bug destruction.
Bad bug reports are usually ignored.
Examples of bad bug reports:
- "Replication server just does't work!"
- "It hangs!!!"
- "Something happened."
Generally, any report with words "some-*" and "just" is useless.
Good bug reports include exact versions of all relevant software: OS,
DBMS servers, DBMS clients, replicator.
How to produce a good report
- Use debug version of ReplServer which is installed into the same directory as the production one.
- Run ReplServer-debug with '--debug' switch. This switch can be set in command line for application mode or in Windows' services management center for service.
Check that debug log is produced - it is much bigger than usual one because it is full of names of called routines.
- Wait till the bug appear.
- Stop Replication server if it is still controllable.
- Send up to 20 last relevant lines from the log to support list (if Replserver "just hang").
- Gather all pieces of log file (if you configured separated logs for pieces of replication schema).
- Pack them using your preferable archiver.
- On request, either send the archive to me or make it downloadable and send me URL.
Don't send the archive to the support mail list because it is big and contain too much information about your database which you don't want to public.
In some cases metadata-only backup of source DB for troublesome schema can be required as well. Read Firebird/Interbase manual about using gbak with -m switch.
PS: All above had been written before I encountered perfect instruction by Simon Tatham.