On 02/27/12 6:34 PM, Adam Bruss wrote:
>
> I’m running Postgresql 9.0.3 on a Windows 7 box. There is one large
> database that gets inserted to thousands of times a day every day. The
> problem is the Handle count on the Windows System process of the
> server, image name c:\windows\system32\ntoskrnl.exe, accumulates over
> time and the handles never get released. This causes the handle count
> to go to about 130,000 at which time the machine starts malfunctioning
> and a reboot is needed. It takes about 6 days before a reboot is
> needed. This is causing our regression test machines, which do inserts
> into the database via a gsoap connection to a web service running on
> the database machine, to get bungled up every 6 days or so. Has anyone
> had a problem with their postgresql server where the System process
> accumulates handles and doesn’t release them?
>
> P.S. In addition to postgresql, the Windows 7 server is running IIS 7
> and asp.net web services. The web services are reached via gSOAP from
> various test machines on the local network, some running windows and
> some running linux.
>
> I really wish I knew what was causing this. I guess it could be
> postgresql, the devart bindings, gsoap or the web service code itself.
> I’m sure I’m closing all database connections in the web service code.
> The database and web services function perfectly except for the handle
> problem.
>
IIS7 runs the websites half in the kernel, which is the 'system'
process... I'd be suspicious of something happening over there, and not
in the postgres database server.
you want to try a test? run postgres on a different system, and have
your webpile connect to it over the LAN. if you're still loosing
handles, its definitely not PG's fault.
--
john r pierce N 37, W 122
santa cruz ca mid-left coast
