rodrigogoncalves wrote:

>
>
>>...the difference between the oldest snapshot
>>transaction and the next is nearly 124,000. That's a lot. Your client
>>has a snapshot transaction lurking around somewhere...
>>
> that interesting... though I'm pretty sure all our code uses only read
> commited as the isolation mode. But I will have a look at it to make sure.
>

It could also be a read-committed, come to think of it. There's a way
to reduce the impact of long-lived read-committed transactions that
we're talking about on the devel list, but it's not in any code yet. So
any long-lived transaction except an explicitly read-only read-committed
transaction could be the problem.


Regards,

Ann