On Thu, May 20, 2010 at 3:36 PM, Tom Lane
<tgl@sss.pgh.pa.us> wrote:
> Setting the breakpoint (b exit) got me a little farther...
> DEBUG: invoking IpcMemoryCreate(size=32595968)
> DEBUG: max_safe_fds = 980, usable_fds = 1000, already_open = 10
> Detaching after fork from child process 19066.
> Detaching after fork from child process 19067.
> Breakpoint 1, 0x0046efb6 in exit () from /lib/libc.so.6
> (gdb) bt
> #0 0x0046efb6 in exit () from /lib/libc.so.6
> #1 0x082286de in proc_exit ()
> #2 0x0820bbdf in ?? ()
> #3 0x0820eb63 in ?? ()
> #4 <signal handler called>
> #5 0x003ec424 in __kernel_vsyscall ()
> #6 0x0051433d in ___newselect_nocancel () from /lib/libc.so.6
> #7 0x0820d489 in ?? ()
> #8 0x08210489 in PostmasterMain ()
> #9 0x081b92df in main ()
Hmm. Not tremendously helpful --- if you were to install the
appropriate postgresql-debuginfo RPM, the stack trace might get
more useful. However, this does appear to confirm the theory
that postmaster.c is doing proc_exit(1), and I don't see any code
paths in there where there isn't a error message logged first.
So right now I'm wondering whether you're not failing to notice
a relevant log message(s). You've evidently managed to launch
the syslogger --- where is it configured to write the postmaster
log?
Gosh, you know I never even bothered looking in pg_log. Stupid! I was expecting to see some sort of error when doing:
$ postgres -d 3 -D /data/postgres/
I even have this written down in my own recovery instructions:
4. Create pg_xlog directory as postgres user, remove backup_label
I had forgotten when doing a backup with WAL enabled you want to exclude the pg_xlog directory because you are restoring from the WAL's written on the standby server in the event of failure. In the pg_log it clearly prints the error.
regards, tom lane
Thanks Tom, I probably would have been running in circles for days without your help.
-Matt
