Le 14:43 06/05/2016,Simon Slavin ?crit: >On 6 May 2016, at 1:32pm, Stephan Buchert <stephanb007 at gmail.com> wrote: > >> The largest database file has now grown to about 180 GB. I need to have >> copies of the files at at least two different places. The databases are >> updated regularly as new data from the satellites become available. >> >> Having the copies of the file synced becomes increasingly tedious >> as their sizes increase. Ideal would be some kind of >> incremental backup/sync facility. > >Believe it or not, the fastest way to synchronise the databases is not to >synchronise the databases. Instead you keep a log of the instructions used to >modify the database. You might, for example, modify the library that you use >for INSERT, DELETE and UPDATE commands to execute those commands and also save >the command to another 'commandLog' table. Or perhaps just append those >commands to a plain text file. > >Then instead of sending any data to the other sites you send this list of >commands to the other sites and have them execute them. > >Once you start implementing this you'll see that it's more complicated than I >have described but the text of your post suggests that you're a good enough >programmer to do it properly.
This does not need to be so complicated: the problem can be solved by three triggers (per table) that insert in a log table the components of the three data modification commands. At definite time points, the contents of this table is used to generate the SQL data modification commands for the other DB. J-L Hainaut
