[clug] Postgres sync / duplication
thestar at fussycoder.id.au
Sun May 7 05:00:49 GMT 2006
Sorry for the top-posting, but there has been a huge amount of recent
work in this.
See the various projects for sync'ing postgresql databases.
Some are proprietary, and some are Free.
>From what I can see, master/slave replication is relatively easy and simple,
but multi-master replication is quite difficult.
Robert Edwards wrote:
>This is such an artificially restricted situation, that it is hard to
>imagine how it would
>apply in real life, but, given the scenario, you could run a job when
>the user logs out
> - inspects the logs on each PostgreSQL server looking for inserts,
>updates and deletes
> - determines in which direction the syncs need to take place
>(checkpoint each log)
> - apply the equivalent inserts, updates and deletes to the other server
>In theory, if both started in sync, then both should remain in sync.
>Dunno what to do
>if something goes wrong, though.
>Just my quick thoughts on the issue.
>Michael Carden wrote:
>>Imagine this scenario:
>>Two servers. One Win2k with NTFS, one Red Hat Enterprise 3 with ext3. Each
>>runs a Postgres server containing content identical to the other. Each has
>>exactly one client workstation connected to it and no connection to anywhere
>>else. Each has one metric buttload (and growing) of RAID attached to it.
>>Imagine this situation:
>>One and only one user needs to do some stuff with one of the databases. User
>>logs into a workstation on either the Win or the Lin system and does work.
>>Now, we need to synch the contents of the database that the user just fiddled
>>with, across to the database on the other system. Nobody else is doing
>>anything with either system once the user has finished, so stopping stuff and
>>restarting is not a problem. The user may have chosen to work on the Win or
>>the Lin database and it should make no difference which. The systems are not
>>normally linked (separate Ethernet networks and separate fibrechannel to
>>storage), but they could be for the duration of the 'synch.'
>>If the Win system went down for a month for some reason and work continued on
>>the Lin system, it should be possible to synch again when the Win system is
>>available again. And same if the Lin system goes down.
>>Any suggestions as to how this might be done?
More information about the linux