The following bug has been logged online:
Bug reference: 6201
Logged by: Jerome Schulteis
Email address: jerome.schulteis@edstrom.com
PostgreSQL version: 9.0.4
Operating system: Windows XP Pro SP3
Description: Windows User Log Off Causes Backend Exception 0xC0000142
Details:
It does not happen on every log off, but if the Windows console user logs
off
of the server at just the wrong time while a backend is starting up, the
backend terminates with exception 0xC0000142 (STATUS_DLL_INIT_FAILED), and
the
PostgreSQL Windows service stops (shared memory block is still in use):
2011-09-09 10:34:01 CDT LOG: server process (PID 3260) was terminated by
exception 0xC0000142
2011-09-09 10:34:01 CDT HINT: See C include file "ntstatus.h" for a
description of the hexadecimal value.
2011-09-09 10:34:01 CDT LOG: terminating any other active server processes
[...]
2011-09-09 10:34:02 CDT WARNING: terminating connection because of crash of
another server process
2011-09-09 10:34:02 CDT DETAIL: The postmaster has commanded this server
process to roll back the current transaction and exit, because another
server process exited abnormally and possibly corrupted shared memory.
2011-09-09 10:34:02 CDT HINT: In a moment you should be able to reconnect
to the database and repeat your command.
2011-09-09 10:34:02 CDT LOG: all server processes terminated;
reinitializing
2011-09-09 10:34:12 CDT FATAL: pre-existing shared memory block is still in
use
2011-09-09 10:34:12 CDT HINT: Check if there are any old server processes
still running, and terminate them.
Originally encountered with our web app running a stress test on PostgreSQL
8.4.2; I reproduced it on a default one-click install of 9.0.4. I run the
following Java on a separate machine to get a new connection every 100 ms:
class ConnectionHog {
static public void main(String[] args) {
try {
while (true) {
java.sql.DriverManager.getConnection(args[0], args[1],
args[2]);
Thread.sleep(Long.parseLong(args[3]));
}
}
catch (Throwable t) {
System.err.println(t);
}
}
}
I then log on and off the server from a separate user account (that is, not
the postgres
account); the error shows up after at most 3 log offs. As long as no log
offs happen on the server, no problem.