Hi Christopher,

Thanks for your reply.

So there is a significant impact. I was sure about that.



________________________________
 De : Christopher Browne <cbbrowne at afilias.info>
À : David TECHER <davidtecher at yahoo.fr> 
Cc : Slony Hackers <slony1-hackers at lists.slony.info> 
Envoyé le : Mercredi 4 avril 2012 21h34
Objet : Re: [Slony1-hackers] Number of lock for CreateEvents and performance
 
On Wed, Apr 4, 2012 at 2:34 PM, David TECHER <davidtecher at yahoo.fr> wrote:
>
> Hi
>
> I am using Slony 1.2.22.  When a createEvents occurs, I noticed that AccessShareLock is assigned on 11700 sequences in the same transaction.
> I've got a lot of objects to replicated.
>
> Could the number of objects being locked having when a createEvent ocurs have a impact?

This isn't a particular surprise...

Upon creating an event, the slon process needs to capture which
sequence values have changed.

In 2.0, we optimized this to only capture those that changed.  In 1.2,
all values of all sequences are captured in sl_seqlog in *every* SYNC
(which corresponds to createEvent).

The improvement in 2.0 means that less data is tracked, though the
locking would still occur on the 11700 sequences.

So all in all, this isn't a phenomenon that I find surprising.

You'll find that table sl_seqlog is a *very* busy table in your
replication infrastructure.  11700 sequences is a mighty lot of
sequences to be replicating.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.slony.info/pipermail/slony1-hackers/attachments/20120404/c3b8498b/attachment.htm