On 9/20/07, Jeff Frost <jeff at frostconsultingllc.com> wrote:
>
> On Thu, 20 Sep 2007, Andrew Hammond wrote:
> > Really? I know that a lot of people, including myself, have _wanted_ a
> slony
> > aware pgpool and for quite some time now. I'm not aware of any such code
> > actually being written. But that's for client software anyway, certainly
> not
> > the slons. Why on earth would anyone even _want_ connect a slon through
> > pgpool?
>
> Sorry, you must be reading serially through your mail.  I believe in a
> later
> email I specified that the slons do not connect through pgpool nor does
> slonik.  pgpool is only being used to load balance read-only search
> queries.
>
> Perhaps by 'slony aware' I should have said 'slony friendly'?  I.e. it
> sends
> SELECTS to both nodes and other queries to the master only.  (Other
> queries
> being defined as non-selects and anything wrapped in a transaction block.)
>

With apologies to H.L. Mencken, that sounds like a simple, direct, plausible
and wrong solution to a complex problem. If pgpool can be configured to send
_transactions_ which have been SET TRANSACTION READ ONLY; to a subscriber
then this might be clever. Otherwise it's not only a stunningly obvious race
condition but total failure to understand locking and mvcc.

Andrew
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.slony.info/pipermail/slony1-general/attachments/20070924/=
173caa31/attachment-0001.htm