On Sep 12, 2007, at 11:37 PM, Peter Wilkinson wrote:
One point to add to that which was part of my motivation for
messing with SQLite as a backend was simple replication of the
database for disaster recovery. Using SQL and its locking seems in
my mind much less prone to problems than something like pausing the
main server and running rsync. Getting reliable and simple
replication going with the standard storage engines is something
I'd like to work on at some point - if anyone has gone down this
path I'd be very interested in the various solutions. I did bring
this up a while ago and got some good pointers but haven't really
had a chance to did into it much.
I think rsync can be used on a Durus database without pausing the
main server.
Here, we just use a cron job with scp from a remote machine to pull a
copy at
regular intervals.
To test the idea, I remember writing a replicator that worked more on a
record by record basis. You just connect to a (possibly
remote) storage server that you want to replicate, and also a local
FileStorage. You iterate on the oid-records of the remote database and
copy those into the new. That establishes the base copy. After that,
you use the invalidation lists to find out what records to re-
retrieve, and
do this in a polling loop, sleeping a little to give the remote server
a chance to do something useful. Even though this is fairly simple,
rsync or scp seems even simpler.
_______________________________________________
QP mailing list
[email protected]
http://mail.mems-exchange.org/mailman/listinfo/qp
