Re: Backend memory dump analysis

On 24 March 2018 at 03:01, Andres Freund <andres(at)anarazel(dot)de> wrote:

> Hi,
>
> On 2018-03-23 14:33:25 -0400, Tom Lane wrote:
> > func_cxt = AllocSetContextCreate(TopMemoryContext,
> > "PL/pgSQL function context",
> > ALLOCSET_DEFAULT_SIZES);
> > plpgsql_compile_tmp_cxt = MemoryContextSwitchTo(func_cxt);
> >
> > function->fn_signature = format_procedure(fcinfo->flinfo->fn_oid);
> > + MemoryContextSetIdentifier(func_cxt, function->fn_signature);
> > function->fn_oid = fcinfo->flinfo->fn_oid;
> > function->fn_xmin = HeapTupleHeaderGetRawXmin(procTup->t_data);
> >
> > This would cost an extra char * field in struct MemoryContextData,
> > which is slightly annoying but it doesn't exactly seem like a killer.
> > Then the memory stats dump code would just need to know to print this
> > field if it isn't NULL.
>
> That's not a bad idea. How about storing a Node* instead of a char*?
> Then we could have MemoryContextStats etc support digging out details
> for a few types, without having to generate strings at runtime.
>

That'd render it pretty useless for extensions, though.

I like the idea of being able to introspect state for particular kinds of
contexts, and not have to generate strings that 99.99% of the time won't
get get looked at.

Function pointers instead of char* ? It adds a significant potential
stability risk to MemoryContextStats() calls, but a great deal of
flexibility.

--
Craig Ringer http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services