On Mon, Aug 28, 2017 at 7:57 AM, Craig Ringer <craig@2ndquadrant.com> wrote:
> I'm frequently debugging postmasters that are around long enough. Deadlocks,
> etc.
>
> It's also way easier to debug shmem related issues with a live postmaster vs
> a core.
Yeah. I don't *frequently* debug postmasters that hang during the
regression tests, but I definitely have done it, and I think something
like this would make it easier. Right now if something wedges and you
need to connect to the postmaster to see what's going on, you have to
grep for the pid, then lsof to get the socket directory. This would
simplify things.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company