Re: hung backends stuck in spinlock heavy endless loop

On Tue, Jan 13, 2015 at 3:54 PM, Andres Freund <andres(at)2ndquadrant(dot)com> wrote:
>> I don't remember seeing _bt_moveright() or _bt_compare() figuring so
>> prominently, where _bt_binsrch() is nowhere to be seen. I can't see a
>> reference to _bt_binsrch() in either profile.
>
> Well, we do a _bt_moveright pretty early on, so that actually might be
> cache misses we're primarily seing.

Still, I see zero references to _bt_binsrch(). Not one. Even if there
was only one non-meta page (i.e. pre-first-root-split), and regardless
of whether this was a read or a write to the B-Tree, there'd still be
some call there as the B-Tree was initially scanned (before _bt_next()
calls, of which I also see none).

--
Peter Geoghegan