HI, The problem with this solution is that i don't have enough space on the targeted live to implement such a procedure. It would have to work by pulling data directly off the live server. The issue with new data whilst the procedure is running (over a span of days or weeks in not severe as each daily activity generates new tables)
As someone else mentioned - I am against implementing any kind of replication as the live server will only have a subset of the data after i'm migrating thus making replication useless. And i also don't have the bandwidth to pull out the data in one day. Regards, Vladislav Geller On Jan 18, 2011, at 8:39 AM, Vladimir Rusinov wrote: > > > On Tue, Jan 18, 2011 at 8:07 AM, Vladislav Geller > <[email protected]> wrote: > Hello, > > I'm currently in the process of migrating a huge live database from one part > of the world to the other (for local use and data analysis). The bandwidth > does not allow me to get a decent transfer speed. Furthermore i can not > migrate during business hours since the connection is critical. This leaves > me with a timeframe of 9 hours a day where i can migrate this database. Does > anyone have experience he is willing to share in respect how to migrate such > databases? > > run pg_start_backup() and use rsync on live data files to transfer them. > If sync won't finish in 9 hours, abort it, run pg_stop_backup() and continue > next day - I assume most of the data won't change, so rsync won't re-transfer > it (but will calculate hash and compare). > > -- > Vladimir Rusinov > http://greenmice.info/
