-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

cbbrowne a écrit :
> Geoffrey wrote:
>> Noticed yesterday that the same two tables on all our databases (12)
>> are out of sync.  I've just noticed today that the date on the slave
>> is 1 hour behind the master.  Is it possible that this time difference
>> could cause any problems with slony replication?
>>
> In what sense are they "out of sync"?
> 
> Slony-I doesn't use timestamps to control its behaviour; it uses
> internal transaction IDs, so this seems rather surprising.
> 
> I'd appreciate it a lot if you could do some poking at it to see if you
> could determine where specifically the problem arose, as  it would be a
> LOT better to be able to give specific diagnosis of "make sure X doesn't
> happen to your system!" rather than merely saying "oh, you shouldn't let
> times get TOO out of sync otherwise there might be some problems..."
> 

Hi,

I jump in this thread, as I'd like to clarify that point.

In documentation (http://www.slony.info/documentation/requirements.html), it is
said that it's a good practice to have ntp ensuring time is OK on every node.

Is it just a good practice, or is there a risk replication could fall if nodes
 don't have the same time ?

What's more, I remember having seen a replication trying endlessly to sync a
slave without success. Every time the slave finished initial synchronisation,
synchronisation was beginning again, as slon thought the node was far away for
the master.

Did I made a false assumption, then? Or could it really be what would happen in
case times are really different between nodes?

Regards,
- --
Stéphane Schildknecht
PostgreSQLFr - http://www.postgresql.fr
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFJ8IU6A+REPKWGI0ERAtBuAJ93PCWI6IdnfHqoDEIlm7qQSGc0SwCeJ8s1
9t8RoFNezC1XJQS/k0yBHbc=
=ZnMF
-----END PGP SIGNATURE-----