On 01/29/14 17:26, Hanselman, Matthew wrote:
> One follow-up: Do you think 2.1.4 is stable in this regard?

I have actually found the bug in 2.2.STABLE that causes this. As Steve
assumed it is a memory corruption inside the log apply trigger.

2.1.x does not have this bug since it doesn't have an apply trigger at
all, but issues individual SQL statements against the updated tables
directly.

The bug can be avoided by setting the apply_cache_size in the slon
config file higher than the number of possible individual apply query
plans during a single SYNC event. This may bloat the memory size of the
backend, used by slon to apply changes, but better that than anything
else. Note that the apply cache size defaults to 50, which is a VERY low
number. Using 1,000 or even 10,000 should not cause a modern DB server
to run out of memory. There can only ever be one DB connection per node
that creates these plans (the local slon's connection). You should be
good bumping that number instead of downgrading to 2.1.

We will have a Slony 2.2.2 release out really soon.

Sorry for the inconvenience.


Regards,
Jan

> 
> Thanks!
> 
> - Matt
> 
> 
> -----Original Message-----
> From: Jan Wieck [mailto:JanWieck at Yahoo.com] 
> Sent: Friday, January 24, 2014 11:36 AM
> To: Hanselman, Matthew; Steve Singer
> Cc: slony1-general at lists.slony.info
> Subject: Re: [Slony1-general] Sync stopped with slony-2.2.1 on data type mismatch
> 
> On 01/24/14 10:03, Hanselman, Matthew wrote:
>> This looks very similar. I'll leave a comment to this effect on the bug. Thanks!
>> 
>> For what it's worth, I tried restarting the slon daemon on both the master & slave. I got a very similar error, but this time it was complaining that "Lost Angeles" wasn't an integer.
> 
> That does indeed look like memory corruption. I will take a very close look at the apply trigger code.
> 
> In the meantime it may mask this error to increase the size of the apply query cache in the slon conf.
> 
> 
> Thanks,
> Jan
> 
> 
>> 
>> - Matt
>> 
>> 
>> -----Original Message-----
>> From: Steve Singer [mailto:ssinger at ca.afilias.info]
>> Sent: Friday, January 24, 2014 9:59 AM
>> To: Jan Wieck
>> Cc: Hanselman, Matthew; slony1-general at lists.slony.info
>> Subject: Re: [Slony1-general] Sync stopped with slony-2.2.1 on data 
>> type mismatch
>> 
>> On 01/23/2014 03:42 PM, Jan Wieck wrote:
>>> On 01/23/14 15:31, Hanselman, Matthew wrote:
>>>> Running Slony 2.2.1 with Postgres 9.2.6. The slave is getting errors 
>>>> trying to apply a change with this error:
>>>>
>>>>
>>>>
>>>> 2014-01-23 14:55:58 EST ERROR  remoteWorkerThread_1_1: error at end 
>>>> of COPY IN: ERROR:  invalid input syntax for integer: "33.7726910000000018"
>>>>
>> 
>> 
>> Could this be the same thing as that is causing
>> http://bugs.slony.info/bugzilla/show_bug.cgi?id=327
>> 
>> (We suspect memory corruption in the apply trigger)
>> 
>> 
>> 
>> 
>>>>
>>>>
>>>> The column in question is a "double precision" on both master & slave.
>>>> No schema change has happened since Slony was deployed.
>>>
>>> What did happen between slony was deployed and this breakage?
>>>
>>>
>>> Jan
>>>
>>>
>>>
>>>>
>>>>
>>>>
>>>> Is this a bug in Slony, or is there something else I can do to troubleshoot?
>>>>
>>>>
>>>>
>>>> - Matt
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Slony1-general mailing list
>>>> Slony1-general at lists.slony.info
>>>> http://lists.slony.info/mailman/listinfo/slony1-general
>>>>
>>>
>>>
>> 
> 
> 
> --
> Anyone who trades liberty for security deserves neither liberty nor security. -- Benjamin Franklin
> 


-- 
Anyone who trades liberty for security deserves neither
liberty nor security. -- Benjamin Franklin