On Wed, Jun 7, 2017 at 4:58 AM, Peter Eisentraut < peter.eisentr...@2ndquadrant.com> wrote:
> On 6/6/17 15:58, Robert Haas wrote: > > The problem with the status quo (after Peter's commit) is that there's > > now nothing at all to identify the logical replication launcher, apart > > from the wait_event field, which is likely to be LogicalLauncherMain > > fairly often if you've got the launcher. I don't personally see why > > we can't simply adopt Tom's original proposal of setting the query > > string to something like "<logical replication launcher>" which, while > > maybe not as elegant as providing some way to override the > > backend_type field, would be almost no work and substantially better > > for v10 than what we've got now. > > The decision was made to add background workers to pg_stat_activity, but > no facility was provided to tell the background workers apart. Is it > now the job of every background worker to invent a hack to populate some > other pg_stat_activity field with some ad hoc information? What about > other logical replication worker types, parallel workers, external > background workers, auto-prewarm? > > I think the bgw_type addition that I proposed nearby would solve this > quite well, but it needs a bit of work. And arguably, it's too late for > PG10, but one could also argue that this is a design fault in the > pg_stat_activity extension that is valid to fix in PG10. > +1. I definitely think it would be a bad idea to put in what basically looks like a workaround into 10, since the new feature was added there. I'd rather have the fix for pg_stat_activity. We used to keep our query state as a text field and that was a bad idea for many reasons. So we moved it to a separate field. Let's not repeat that mistake here. -- Magnus Hagander Me: https://www.hagander.net/ <http://www.hagander.net/> Work: https://www.redpill-linpro.com/ <http://www.redpill-linpro.com/>
