Re: BUG #2419: could not reattach to shared memory

Hi,

Thank you for your replies.

I accept that the "Permission denied" problem does suggest that the DB error
may be caused by the OS somehow.

There is no problem with the permissions/ownership of the files because the
Postgres account created and owns and those files; this rules out any sort
of security problem. It is possible that the file is inaccessible through
some other reason and that Postgres is merely reporting that it can't access
the file and, that it 'could' be caused by a permissions problem rather than
it 'is' a permissions problem. The error log information in this case isn't
really very useful since it doesn't accurately report the real cause of the
error. The OS doesn't report any other errors and there are no other
systems problem or file access problems; the only problem lies within the
database.

To try and reproduce the problem on another machine, I did a new install of
the same version of Postgres (8.1.3) and dump/restored the database onto
this new server. So far it has been running with the same load and activity
for almost 30 hours and the problem has not surfaced. In theory, Postgres
and the database are identical and therefore, the fact that it doesn't error
in the same way does confirm this is an OS problem (assuming the problem
doesn't occur at some point in the future). The two servers are identical
hardware and have the same version of OS, Windows Server 2003 SP1.

None of this helps me because I still have a production server on which I
can't run the database since I can't debug the error. Any suggestions?

Thank you for your assistance.

Andy

"Andy" <andy(at)otelex(dot)co(dot)uk> writes:
> To try and reproduce the problem on another machine, I did a new install of
> the same version of Postgres (8.1.3) and dump/restored the database onto
> this new server. So far it has been running with the same load and activity
> for almost 30 hours and the problem has not surfaced. In theory, Postgres
> and the database are identical and therefore, the fact that it doesn't error
> in the same way does confirm this is an OS problem (assuming the problem
> doesn't occur at some point in the future). The two servers are identical
> hardware and have the same version of OS, Windows Server 2003 SP1.

We've seen reports of intermittent permission failures on Windows being
caused by broken antivirus software. What security software have you
got on those machines, and does the failure go away if you remove it?

regards, tom lane