For the longest time I've been under the mistaken impression that it was OK to
copy entire __repository__ directories from machine to machine and expect all
to be fine. Not so !
The Berkeley DB documentation is pretty clear about it:
- Environment files are very specific to the machine, architecture and even
OS version they're created on. These are the db.* files in __repository__.
- Log files are byte order specific. These are the log.* files.
- Database file, only, are actually portable. These are the *.db files.
In order to copy a repository around machines of different OSs, or different
byte order:
- if the repository wasn't closed properly, running db_checkpoint first
would ensure that the latest transaction(s) are committed to the actual
database files (on clean close(), checkpoint() is run):
. on Windows:
- make sure $CHANDLERBIN/release/bin is on the PATH (watch out, cygwin
has really old db_* utilities that don't work with our stuff)
- $CHANDLERBIN/release/bin/db/db_checkpoint -1 -h __repository__
. on Unix
- make sure $CHANDLERBIN/release/lib is on the LD_LIBRARY_PATH or
DYLD_LIBRARY_PATH
- $CHANDLERBIN/release/db/bin/db_checkpoint -1 -h __repository__
yes, in one case it's bin/db, in the other db/bin, don't ask...
- copy the db.* files only
To restore a copy:
- create an environment by creating an empty repository or use an existing
environment by deleting the log.* files first
- copy the db.* files into the environment
I expect to write a script soon to simplify this mess...
Andi..
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "Dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/dev
