I asked this on the slony list, but maybe more appropriate to ask here...
I'm using slony to replicate a database with 200 sequences. To replicate
these with slony or our modified dbmirror, the replicator polls every
sequence relation (select last_value...) on each syncronization interval.
That amounts to 200 select queries per sync interval. That can be a
significant load in our case if we want short sync intervals of 1-2
seconds. Naturally, this leads to the idea of triggers on sequences, which
I know do not exist at present in pgsql.
Are triggers on sequences a reasonable feature request/hope for Pgsql?
Short of simply increasing the sync interval, what other options are there
for addressing the performance load in a case like this?
Ed
