Thread: postmaster shutdown - LWLockAcquire?

Hello!

This behaviour is starting to get on my nerves...

Could somebody give me a hint as to what exactly happened here and how I
can avoid it in the future?

This is the relevant bit of the Logfile:
----------------snip----------------
2002-03-06 16:02:52 FATAL 1:  Database "template0" is not currently
accepting connections
2002-03-06 16:02:53 FATAL 1:  Database "template0" is not currently
accepting connections
2002-03-06 16:02:53 FATAL 1:  LWLockAcquire: can't wait without a PROC
structure
2002-03-06 16:02:53 DEBUG:  server process (pid 1848) exited with exit
code 1
2002-03-06 16:02:53 DEBUG:  terminating any other active server
processes
2002-03-06 16:02:53 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 terminate your database system connection and exit.
    Please reconnect to the database system and repeat your query.
[this message is repeated six times]
2002-03-06 16:02:56 DEBUG:  all server processes terminated;
reinitializing shared memory and semaphores
IpcMemoryCreate: shmget(key=5432001, size=2039808, 03600) failed: Not
enough memory

This error usually means that PostgreSQL's request for a shared
memory segment exceeded available memory or swap space.
To reduce the request size (currently 2039808 bytes), reduce
PostgreSQL's shared_buffers parameter (currently 128) and/or
its max_connections parameter (currently 32).

The PostgreSQL Administrator's Guide contains more information about
shared memory configuration. [that wasn't really too helpful on this
matter; next I started up the service again. The subsequent entires go
along as follows:]

2002-03-06 16:15:25 DEBUG:  database system was interrupted at
2002-03-06 15:58:24
2002-03-06 16:15:25 DEBUG:  checkpoint record is at 1/E3952A78
2002-03-06 16:15:25 DEBUG:  redo record is at 1/E3952A78; undo record is
at 0/0; shutdown FALSE
2002-03-06 16:15:25 DEBUG:  next transaction id: 8343789; next oid:
2876559
2002-03-06 16:15:25 DEBUG:  database system was not properly shut down;
automatic recovery in progress
2002-03-06 16:15:25 DEBUG:  redo starts at 1/E3952AB8
2002-03-06 16:15:25 DEBUG:  ReadRecord: record with zero length at
1/E39594F8
2002-03-06 16:15:25 DEBUG:  redo done at 1/E39594D0
2002-03-06 16:15:27 DEBUG:  database system is ready
----------------snip----------------

I have absolutely no idea what I should do about this. What does
"LWLockAcquire: can't wait without a PROC structure mean?" Most of the
time the thing is running fine, but it just quits occasionally. It's
PostgreSQL 7.2 running on Cygwin, Win2k, 2GB RAM, quad-processor by the
way, we are doing access via ODBC. I'd be happy to provide any more
information if it could help in solving the problem - and I'd be very
happy about helpful advice.

Thanks in advance,

   Markus

"Markus Wollny" <Markus.Wollny@computec.de> writes:
> This is the relevant bit of the Logfile:

What are your client(s) doing to provoke this crash?  (If you don't
already know, try running the postmaster at -d2 to get logging of
queries.  Might want to turn on log_connections and log_pid too.)

I doubt that the template0 connection attempts are causing the crash;
at least that doesn't seem to cause any problem here.

The LWLockAcquire failure suggests that some code path is trying to
invoke LWLockAcquire before locking support has been set up in a new
backend.  But I have no idea where that could be, if it's being seen
by you and not anyone else.

> PostgreSQL 7.2 running on Cygwin, Win2k, 2GB RAM, quad-processor by the
> way, we are doing access via ODBC. I'd be happy to provide any more
> information if it could help in solving the problem - and I'd be very
> happy about helpful advice.

I don't suppose you'd consider "run a real OS" as helpful advice ;-)

            regards, tom lane