On 7/16/2007 8:56 AM, Bill Moran wrote:
> In response to "Al Rosenthal" <arosenthal at AtlantaHand.com>:
> 
>> I have thought about firewalls and timeouts.  I just have two machines, 
>> sitting next to each other connected by a switch.  One has Vista, the other 
>> XP.  I tried running the slon engines on either one.  I also tried changing 
>> all of the IP settings available.  I also tried adding a linux machine and 
>> running the slon process on that machine.  In every case, the machine that 
>> is the slave failed after about 12 - 15 GB with a error 10055 (not enough 
>> buffer space or enough tcp buffers).
> 
> Perhaps try bumping up slon's log level and try to correlate what it's doing
> when the problem occurs.

That's what I'd like to know. What exect operation slon is doing is 
actually failing (if it isn't actually Postgres that is getting that 
specific error).

Also interesting would be the output of "netstat -na" at the time of the 
failure. Is there anything else going on on the systems while the slons 
are creating the replica? Any anti virus software that might mess with 
the network? Have you tried to disable useless services like indexing 
and such?

> 
>> Also, the database I am trying to 
>> replicate is very simple.  One table, no indexes.  Even the table is simple. 
>> On serial column as primary key, the other is bytea data.
> 
> Note that Slony can not replicate bytea fields.

When and by whom was that broken? As per design, Slony cannot replicate 
Postgres' large objects, but bytea should be just fine.


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 #