I'm sorry, but the issue has been fixed now; we've upgraded to postgres-8.3, so I can't reproduce this anymore..
I'm not completely sure how it got fixed (it wasn't me), but I can ask the person who fixed it and get back to you
tomorrow;perhaps that might help narrow the field a little bit.
On Tue, Nov 10, 2009 at 09:31:06AM -0500, Tom Lane wrote:
>"Helge Milde" <helge@monsternett.no> writes:
>> It seems the recovery process is failing; doing a strace on the 'postgres:
>> startup process' pid, I see this:
>
>> -- strace log start ---
>> Process 29888 attached - interrupt to quit
>> close(14) = 0
>> open("pg_clog/021B", O_RDWR|O_CREAT|O_LARGEFILE, 0600) = 14
>> _llseek(14, 245760, [245760], SEEK_SET) = 0
>> write(14, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 8192)
>> = 8192
>> fsync(14) = 0
>> close(14) = 0
>> open("pg_clog/021B", O_RDWR|O_CREAT|O_LARGEFILE, 0600) = 14
>> _llseek(14, 245760, [245760], SEEK_SET) = 0
>> write(14, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 8192)
>> = 8192
>> fsync(14) = 0
>> close(14) = 0
>> -- strace log end ---
>
>> This repeats infinetly.
>
>Hmm, can you attach to that process with gdb and collect a few stack
>traces to show where it's looping? This isn't a symptom we've seen
>before, AFAIR.
>
> regards, tom lane
--
Helge Milde, 69701808
www.monsternett.no