"surabhi.ahuja" <surabhi.ahuja@iiitb.ac.in> writes:
> <2006-05-24 16:21:28 CDT%>LOG: database system was shut down at =
> 2006-05-24 16:01:00 CDT
> <2006-05-24 16:21:28 CDT%>LOG: record with zero length at 0/AC92BC
> <2006-05-24 16:21:28 CDT%>LOG: invalid primary checkpoint record
> <2006-05-24 16:21:28 CDT%>LOG: record with zero length at 0/AC9280
> <2006-05-24 16:21:28 CDT%>LOG: invalid secondary checkpoint record
> <2006-05-24 16:21:28 CDT%>PANIC: could not locate a valid checkpoint =
> record
> <2006-05-24 16:21:28 CDT%>LOG: startup process (PID 8428) was =
> terminated by signal 6
Hm, those are remarkably small values of the checkpoint record pointers;
roughly what you'd see immediately after initdb. What's the history of
this installation? If it's not freshly initdb'd, the only thing I can
think is that somebody has restored an old version of pg_control into
the directory tree. Which would be a pretty bad move, but maybe you can
recover with pg_resetxlog.
regards, tom lane
