Sebastian Arcus wrote: > On 12/02/16 20:31, Antony Stone wrote: >> On Friday 12 February 2016 at 17:29:23, Sebastian Arcus wrote: >> >>> As per advice from this list, I have been re-using my bayes databases on >>> several different servers running SA. On one of the servers though, the >>> database is not accepted. >> Are the servers all the same distro, release and version? >> >> Are they all up to date (or at least, all equally up to date as each >> other)? > No - there are definitely variations in kernel versions and other > software versions. So far I have only made sure the SA version is the > same. Any possible culprit comes to mind in terms of a certain library > or software that might cause this?
Standard on-disk Bayes files are (mostly) BerkeleyDB, but there are a couple of other possible back ends that could be in use depending on OS release and update level. If you have OS variations you may have different BDB versions (or even a completely different type of hash database file; look up AnyDBM_File), and you won't be able to share your Bayes by copying files around. Try running "file /path/to/bayes/files/*" on your source and problem machines (the latter with locally-created Bayes files that don't show the access problem) and see what comes up. If updating the problem machine (or updating all the rest of them, if the problem machine is "newer") is a problem for any reason, you might try sa-learn --dump to "back up" Bayes on your source machine, then sa-learn --restore to reload the data on another machine. In general though, if you're operating at a scale where one server isn't enough to handle your SA load, you may want to start thinking about SQL for Bayes, which can be shared much more easily than pushing file-based Bayes data around. -kgd
