On Feb 8, 2005, at 12:55 PM, Christopher Browne wrote:

>> Performance is again excellent, if not better than before.  Perhaps a 
>> vacuum full of the slony tables after the initial copy is a good 
>> idea?
>>
>
> VACUUM FULL of the tables where???  On the origin?  Or on the 
> subscriber?
>

The subscriber.  The origin seems to pretty much be under control.  I 
ran vacuum full on the sl_log_1 there too, but there weren't many rows 
to reclaim.  basically it wasn't worth it :-)

> I'd be comfortable enough doing it on a subscriber that is still 
> catching up and therefore isn't being used for production purposes...
>
> If you'd been running generate_syncs.sh, it would have come up to date 
> in a much "friendlier" fashion, and not needed the vacuum nearly so 
> badly.  It probably would have been able to do enough cleanup while 
> getting up to date that the FSM wouldn't have been blown out...
>

Yes, had the sync's been generated it would not have been such an 
issue.  The table bloat did kill my performance for several days, 
unfortunately.

Does generate_syncs work with 1.0.5?  I remember something about it 
being for -HEAD.