On Fri, 20 Jan 2012, Zac Bentley wrote:

> We have a test cluster of machines that we run Slony
> (2.0.7/postgres8.4/centos6) on in our facility. In the past, in order to
> update the cluster to the latest application code, we have simply wiped the
> servers, reinstalled a fresh configuration of centOS, Slony, postgres etc,
> restored the same pg_dump to each database, and then created the main
> replication set and kicked off subscription. This worked for the past few
> months, but recently, every time I kick off initial subscription, the
> postgres database files simply grow until the servers run out of disk space.
> This is more than a doubling of our database size; it is going from 18G
> (normal) to around 40G. I can watch the Slony logs, and see that a table
> copy on a very large and complex table in our application has just finished
> (the table alone is 2GB), and I can watch the postgres server status and see
> that FinishTableAfterCopy is running on the large table. However, the server
> always runs out of disk space before subscription can complete. Our
> application has experienced slow linear growth for the past months, and none
> of these tables are unusually large. Is there any reason you can think of
> that subscription has started filling up 2x the database's size on the disk?

It isn't clear to me from your description if the slave is growing or if the 
master is growing.  Also check which tables are growing/using up the space?
(pg_size_pretty(pg_total_relation_size())

While the subscription is going on sl_log_1 or sl_log_2 on the master will 
grow until the entire subscribe process is complete but you would have to be 
doing a significant transaction volume to double your database size that 
way.

Steve