I've resolved this issue. But I had to ship the databases to the remote
location and reload them again to do it. I inherited this system and we
are working on updating it to postgesql 9.3, but here's where we are:

Master - postgres 8.3.9, ubuntu 8.04.4, slony 2.0.3 RC2
Slave1 - postgres 8.3.7, ubuntu 8.04.3, slony 2.0.3 RC2

Slave1 cascades to Remote1, using log shipping (starting slon with -a
option)
Remote1 - Postgres 9.2.4, RHEL 5.9, slony1-92-2.1.3-1.rhel5

The problem occurred when Slave1 crashed due to a hardware failure. There
were a few logs created with 0 length, but correct sequence number.
Remote1 couldn't process those log files, they have no sql inside.

So, in the end, I got the databases on Slave1 back in sync with Master,
then dumped them and sent the data to Remote1, where I reload the
databases and restarted log shipping.

Mike

On 9/10/13 7:23 PM, "Steve Singer" <steve at ssinger.info> wrote:

>On Tue, 10 Sep 2013, Mike James wrote:
>
>> I have been sending slony1 logs over a WAN to replicate to a remote
>>server. A couple of days ago, the slave
>> that was sending logs crashed. It was rebooted successfully a couple of
>>hours later, but it looks like log
>> shipping did not resume.
>> 
>> There are a few logs with 0 length on the provider, but they can't be
>>consumed by the remote server on the
>> other end of the WAN.
>> 
>> Is there any way to recover gracefully from this outage?
>
>What version of slony are you running, and does slony_logshipper exit
>immediately after starting up  or stay running and but does nothing? Does
>it 
>print anything in its logfile?
>
>
>> Thanks, Mike
>> 
>>
>