On 11/11/2004 4:40 PM, Rod Taylor wrote:

> On Thu, 2004-11-11 at 16:37 -0500, Jan Wieck wrote:
>> On 11/11/2004 3:35 PM, Rod Taylor wrote:
>> 
>> >> It seems that during the transfer of data from one system to the other
>> >> the slon process is slowly growing (730MB in size, 32MB resident) at
>> >> this time. Since I'm down to 300MB of swap free out of 4.8GB total
>> >> memory,  with PostgreSQL using around 3GB, I think I've found the source
>> >> of the crash -- out of memory.
>> > 
>> > I should mention that the disk isn't thrashing memory in and out, so I
>> > don't think that most of the memory is actively used.
>> 
>> ... which supports the theory of a leak ... (not that I have enough 
>> information to go hunting).
> 
> I've got a backup of the master (source) from about mid-way through the
> slon process, including all of the slony log data. I don't have a copy
> of the slave at the same point.
> 
> If you can tell me how to use a newly restored master to go through the
> process again (and how to get a profile or the information needed), I
> can probably dig up the 1/2TB of diskspace required to give it a try.
> 

The thing is, I have used valgrind to get memory leaks out of slon as 
far as I can. And if it is a general memory leak, the long-term runs I 
am doing here at home should show "something" at least. Is it possible 
that this employs a memory leak inside of the 7.2 libpq? Gotta check the 
7.3 and 7.4 release notes ...


Jan

-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #