Since the beginning of the week, we get strange errors in our event
log:

[sqsrvres] CheckQueryProcessorAlive: sqlexecdirect failed
[sqsrvres] printODBCError: sqlstate = 08S01; native error = 2746;
message = [Microsoft][SQL Native Client]TCP Provider: An existing
connection was forcibly closed by the remote host.
[sqsrvres] printODBCError: sqlstate = 08S01; native error = 2746;
message = [Microsoft][SQL Native Client]Communication link failure
[sqsrvres] OnlineThread: QP is not online.
[sqsrvres] CheckQueryProcessorAlive: sqlexecdirect failed
[sqsrvres] printODBCError: sqlstate = 08S01; native error = 0; message
= [Microsoft][SQL Native Client]Communication link failure
[sqsrvres] CheckQueryProcessorAlive: sqlexecdirect failed
[sqsrvres] printODBCError: sqlstate = 08S01; native error = 0; message
= [Microsoft][SQL Native Client]Communication link failure
[sqsrvres] CheckQueryProcessorAlive: sqlexecdirect failed
[sqsrvres] printODBCError: sqlstate = 08S01; native error = 0; message
= [Microsoft][SQL Native Client]Communication link failure
[sqsrvres] CheckQueryProcessorAlive: sqlexecdirect failed
[sqsrvres] printODBCError: sqlstate = 08S01; native error = 0; message
= [Microsoft][SQL Native Client]Communication link failure
[sqsrvres] CheckQueryProcessorAlive: sqlexecdirect failed
[sqsrvres] printODBCError: sqlstate = 08S01; native error = 0; message
= [Microsoft][SQL Native Client]Communication link failure
Login failed for user 'xxxuser'. [CLIENT: xx.xx.xx.xx]
[sqsrvres] CheckQueryProcessorAlive: sqlexecdirect failed
[sqsrvres] printODBCError: sqlstate = 08S01; native error = 2746;
message = [Microsoft][SQL Native Client]TCP Provider: An existing
connection was forcibly closed by the remote host.
...

It has done this 2 times. The first time, the sql server service
stopped and restarted. The second time there was no restart.
The cluster has an active/active configuration. It is always the second
instance that gives the errors.  I think it has something to do with
memory consumption.
On node 1 there are 20 databases, on node 2 there is only one.
But when I have a look into the cached plans, I can see that the total
of cached plans on instance1 is around 20000 and on instance2 it is
around 40000.

Our cluster consists of two nodes with each about 8GB of RAM. There are
2 AMD Opteron Processors inside of each node.
The VM size, I've set it on an initial size of 6GB with a maximum of 10
GB on both nodes.
SQL server is configured to use a maximum of 6GB on both nodes.
OS installed:Microsoft Windows Server R2, Enterprise x64 Edition,
Service Pack 1

I don't know if it is a coincidence, but the previous week, they did a
windows update on both nodes (express update). Following has been
installed:
Security Update for Microsoft .NET Framework, Version 2.0 (KB917283)
Security Update for Windows Server 2003 x64 Edition (KB921398)
Security Update for Outlook Express for Windows Server 2003 x64 Edition
(KB920214)
Security Update for Windows Server 2003 x64 Edition (KB917422)
Security Update for Windows Server 2003 x64 Edition (KB922616)
Security Update for Windows Server 2003 x64 Edition (KB920683)
Security Update for Windows Server 2003 x64 Edition (KB920670)
Cumulative Security Update for Internet Explorer for Windows Server
2003 x64 Edition (KB918899)
Security Update for Windows Server 2003 x64 Edition (KB921883)
Security Update for Windows Server 2003 x64 Edition (KB917537)

All clients have been patched too with the latest security updates.

The sqlservr process now uses 3,6GB of memory, with a peak of 5,6GB, VM
size is 6.3GB.

If anyone should have a solution or tips for that, it would be very
helpfull.

Kind Regards
Geoffrey