Is there an easy way to get a count on the number of transactions postgresis processing a min/hour/day/etc.? If so, can it be broken down by theindividual databases in the cluster?
Thanks,
Chris