Attached to backend postmaster and got the following. Hope this helps.
Attaching to program: /nfs/silence/bigdisk/eurrow/pgsql/bin/postmaster, process 170422
[Switching to Thread 1]
0x000000000000377c in ?? ()
(gdb) break errfinish
Breakpoint 1 at 0x1000019dc
(gdb) cont
Continuing.
[Switching to Thread 1]
Breakpoint 1, 0x00000001000019dc in errfinish ()
(gdb) bt
#0 0x00000001000019dc in errfinish ()
#1 0x00000001002920d0 in reaper ()
#2 <signal handler called>
#3 0x0fffffffffffd810 in ?? ()
Cannot access memory at address 0x203fe94000000000
(gdb) cont
Continuing.
Breakpoint 1, 0x00000001000019dc in errfinish ()
(gdb) bt
#0 0x00000001000019dc in errfinish ()
#1 0x0000000100292680 in LogChildExit ()
#2 0x00000001002971a8 in CleanupBackend ()
#3 0x00000001002923d0 in reaper ()
#4 <signal handler called>
#5 0x0fffffffffffd810 in ?? ()
Cannot access memory at address 0x203fe94000000000
(gdb) cont
Continuing.
-----Original Message-----
From: Tom Lane [mailto:tgl@sss.pgh.pa.us]
Sent: Tuesday, November 14, 2006 11:01 AM
To: Rob Owen
Cc: pgsql-general@postgresql.org
Subject: Re: [GENERAL] Out of memory (Failed on request size 24)
"Rob Owen" <Rob.Owen@sas.com> writes:
> PostgreSQL 8.0.3 running on AIX 5.3 (same thing happens on 5.1 though).
> DBMS was running fine for some months but now one of the databases isn't accessible. Any help would be greatly
appreciated.
Just one database? Sounds like it might be corrupt data in that database's system catalogs. Can you get a stack trace
fromthe point of the error to help us narrow it down?
The way I usually debug startup-time failures is:
export PGOPTIONS="-W 30"
psql ...
Now I have 30 seconds to identify the PID of the backend process in another window and do (as the postgres user)
gdb /path/to/postgres PID
Once you've got gdb control of the backend, do
gdb> break errfinish
gdb> cont
... wait for the timeout to finish elapsing, if needed ... Once gdb reports that the breakpoint has been reached, say
gdb> bt
... useful info here...
gdb> cont
regards, tom lane