Hi everybody,

we had a strange problem at a customer site, which raises some questions here.

The system runs on a Win2003 SP1 2 node cluster with 2 Named Instances of
SQL 2000 SP4 (One instance on each node). The admins planned to install
updates and patches on the nodes and tried to move the second SQL Instance
from the second node to the first node.

On this first node, the SQL Server Resource failed to start and moved back
to node 2. (On node 2 the SQL Server Agent Resource crashed, but this seems
to be another story;-)

We tried to bring the SQL Server Resource online manually on node 1, but
without success. We tried to figure it out and came to the following results:

The SQL Server Resource failed to open Port 1209. The Server is configured
to use dynamic port allocation, but it seems that this didn't worked:
So it started only using Named Pipes and Shared Mem support.
The recorvery of SQL Server completed.
The Cluster Management Extension for SQL Server sqsrvres failed to login to
sql server and logged odbc errors (because the client is only configured to
use TCP/IP (or the Extension uses TCP/IP as a default because it is the only
protocol support by MSCS)):
After 180 Seconds (cluster pending timeout) of repeated and failed
connection attempts the cluster initiated a fall back to the other node.
Now my questions:

1. Does anybody now, how the dynamic port allocation worked in detail?
2. How is it possible that the sql server does not get an free port ?
3. How does the cluster management extension for sql server (sqsrvres)
communicate with the server, is it using all configued methods from Network
Client Utility or does it communicate hardcoded with TCP/IP


Any Help is welcome....

Kind Regards