COM on windows 2003 cluster
- From: Igor.Akopov@xxxxxxxxx
- Date: 19 Oct 2006 07:27:25 -0700
Are there any special considerations or settings when running COM
server on a cluster, compared to single computer? I didn't find
anything on docs or MSDN - but if there's an article on such matters, a
link would be very helpful.
---
We have a setup consisting of COM server and 2 local COM clients: one
is our own (so we have full access to source), another is commercial
(OpcLink). Server and both clients are running as services, under same
user account. COM object is also configured (with dcomcnfg) to run
under the same user account. COM clients connect to 'localhost' with
CoCreateInstanceEx.
It works without any problems on a single machine Windows Server 2003.
But when we deploy it on a two-node active/passive cluster (all 3
services in the same cluster group), a weird problem appears. That is,
OpcLink fails to connect to COM server after our own client:
1. OpcLink connects (success), then our client connects (success)
2. Our client connects (success), then OpcLink connects (failure)
3. OpcLink connects (success), then our client connects (success),
OpcLink disconnects, then connects (failure).
There are no errors in event logs when connection fails.
It looks a bit like it tries to launch new instance of COM server
process if the current process was 'spoiled' by another client (and
fails, because second instance is not allowed).
Services and DCOM are configured exactly same on a single pc and
cluster systems, so I'm pretty sure the problem lies in clustering.
.
- Prev by Date: Re: virtual IP problem
- Next by Date: Re: error 1130 - trying to add 3rd node to cluster
- Previous by thread: What can NLB Handle traffic wise?
- Next by thread: Re: error 1130 - trying to add 3rd node to cluster
- Index(es):
Relevant Pages
|