<div dir="ltr">Thanks for the response.<div><br></div><div>How about when you change the db schema such as adding/removing/renaming tables, columns and dependencies? We usually stop the replication, delete the cluster from both Master and Slave, make the changes on both Master and Slave and run the setup from scratch. This would also freeze postgres and take several hours for replication to start. Isn't there a way to make it faster considering that all data is already in the Slave?</div><div><br></div><div>Also, how do you subscribe with no copy? </div><div><br></div><div><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Sat, Oct 10, 2015 at 12:50 AM, Scott Marlowe <span dir="ltr"><<a href="mailto:scott.marlowe@gmail.com" target="_blank">scott.marlowe@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">On Fri, Oct 9, 2015 at 9:34 PM, Sung Hsin Lei <<a href="mailto:sungh.lei@gmail.com">sungh.lei@gmail.com</a>> wrote:<br>
> Hello,<br>
><br>
> I have a 10gig database on the Master which I backed up and restored on the<br>
> Slave. Database activity is low. When I setup Slony replication from the<br>
> Master to the Slave, it would take hours before new information from the<br>
> Master would be updated into the Slave. Postgres would be also VERY slow for<br>
> the first few hours. However, after the initial wait period, the Slave would<br>
> update within a second or so and Postgres speed would return to normal. I'm<br>
> assuming that Slony is taking time to verify the initial database integrity<br>
> between the Master and the Slave. Am i right? I would like to know if<br>
> there's a way to have the updates start within the first few minutes of the<br>
> original setup.<br>
<br>
</span>In a normal slony sub, all you need on the slave is the schema. If<br>
there's data there it'll get truncated / deleted, so there's no reason<br>
to spend time copying it over.<br>
<br>
IF and only IF you are willing to take your master offline from your<br>
application (i.e. stop writes going to it) you can copy over the data<br>
and then subscribe with no copy. If the master is not taken "offline"<br>
from an application perspective then a subscribe with no copy will<br>
result in your master and slave not having the same data (which is<br>
bad).<br>
</blockquote></div><br></div>