"Steve Wolfe" <steve@iboats.com> writes:
>> Show us a gdb backtrace from the backend core dump?
> Sounds like a good suggestion to me. : )
> Where would the core file be found?
Should be in $PGDATA/base/YOURDB/core.
If it's not there, then either (a) the backend is not actually crashing,
but doing a semi-controlled exit --- check the postmaster log to find
out; or (b) no core dump is being made because you started the
postmaster with ulimit settings that forbid a core dump. (b) is pretty
likely on Linux machines because "ulimit -c 0" is the standard context
for system boot scripts. Try restarting the postmaster with "ulimit -c
unlimited" (I'd suggest putting this into the PG startup script).
regards, tom lane