Thread: 7.1.3 XLogWrite Error

Hi,

I am getting "FATAL 2:  XLogWrite: write request is past end
of log" erros while inserting new records to
a 7.1.3 database. There is sufficient disk space so
I am wondering what might be causing it. The application
selects old data from a 7.1.2 database, processes it
and then add it to the 7.1.3 database. The application
runs fine for a while adding about 70,000 records and
then it breaks. Any ideas what might be causing it.
There is a commit which occurs every 50 record.

Here is the full error text:.


FATAL 2:  XLogWrite: write request is past end of log
Server process (pid 11551) exited with status 512 at Tue Dec
11 22:42:20 2001
Terminating any active server processes...
NOTICE:  Message from PostgreSQL backend:
        The Postmaster has informed me that some other
backend  died abnormally
and possibly corrupted shared memory.
        I have rolled back the current transaction and am
going to termina
te your database system connection and exit.
        Please reconnect to the database system and repeat
your query.
Server processes were terminated at Tue Dec 11 22:42:20 2001
Reinitializing shared memory and semaphores
DEBUG:  database system was interrupted at 2001-12-11
22:40:05 GMT
DEBUG:  ReadRecord: invalid magic number 0000 in logfile 0
seg 5 off 16196284
DEBUG:  Invalid primary checkPoint record
DEBUG:  ReadRecord: invalid magic number 0000 in logfile 0
seg 5 off 9966944
DEBUG:  Invalid secondary checkPoint record
FATAL 2:  Unable to locate a valid CheckPoint record
postmaster: Startup proc 11552 exited with status 512 -
abort
postmaster: Startup proc 11552 exited with status 512 -
abort

Any pointers will be helpful.

Thanks..
Amin

<amin@quantiva.com> writes:
> I am getting "FATAL 2:  XLogWrite: write request is past end
> of log" erros while inserting new records to
> a 7.1.3 database.

Most curious.  We've seen reports of such errors when trying to
restart after a system crash, but not during normal operation.
Would you be willing to poke into it with a debugger, or let
someone who knows the code do so?

            regards, tom lane