Re: Allow "snapshot too old" error, to prevent bloat

On Sun, Feb 15, 2015 at 11:36:50AM -0500, Tom Lane wrote:
> I wonder if we couldn't achieve largely the same positive effects through
> adding a simple transaction-level timeout option.  That would be far
> easier for users to understand and manage, it would be trivial to allow
> specific high-value transactions to run with a laxer setting, it does not
> create any implementation-detail-dependent behavior that we'd be having to
> explain to users forevermore, and (not incidentally) it would be a lot
> simpler and more trustworthy to implement.  There's no well-defined
> correlation between your setting and the net effect on database bloat,
> but that's true with the "snapshot too old" approach as well.

While we have prided ourselves on not generating query cancel messages
due to snapshot-too-old, there is the downside of bloat.  I understand
the need to allow users to make the trade-off between bloat and query
cancellation.

It seems we already have a mechanism in place that allows tuning of
query cancel on standbys vs. preventing standby queries from seeing old
data, specifically
max_standby_streaming_delay/max_standby_archive_delay.  We obsessed
about how users were going to react to these odd variables, but there
has been little negative feedback.

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + Everyone has their own god. +