Bill,

Definitely put those settings that NetApp recommends in place. The oplocks could be a problem. I would also do the following on the filer:

1). priv set diag

2). setflag smb_trace_connect 1 (is not set after a reboot)

3). setflag tcp_trace_RST 1 (is not set after a reboot)

4). printflag (to verify it's set to 1)

5). options rsh.initial_priv_from_console on

(this will allow you to trace rsh to the filer and execute the more privileged filer commands.)

6). priv set admin

7). In the /etc/rc file add the following lines.

priv set diag; setflag smb_trace_connect 1 priv set diag; setflag tcp_trace_RST 1 options rsh.initial_priv_from_console on

These settings will log any TCP resets that get sent from the filer. My own experience is that most of the database drops occur when the filer sends out a RST (reset) command to the database engine. If you have the database go down again, match up the time and see if you do not have a corresponding RST in the messages log. It's a first step.

Also, do you have clustering enabled between two filers?

Bryan Bryan Leonard BancTec, Inc. bryan.leonard@banctec.com

-----Original Message----- From: Bill Baker [mailto:BillBaker@bftg.com] Sent: Tuesday, November 12, 2002 1:22 PM To: toasters@mathworks.com Subject: SQL disconnects from Filer

We are experiencing a problem where our SQL servers lose connectivity to the databases on the Filer. The SQL server application logs in the event viewer show the disconnect but there is no indication of any TCP errors in the System event log. I am not seeing any packet or port errors on the switch these devices connect to. We also have Lotus Notes servers attached to the Filer via the same switch and they do not lose connectivity to their databases.

Network Appliances recommends Windows Registry changes for MaxMpxCt, TcpWindowSize, and UseOpportunisticLocking along with corresponding CIFS options on the Filer. I discovered our SQL servers do not have these Registry entries and the CIFS option for oplocks on the Filer is set to on rather than off.

Would this configuration contribute to the loss of connectivity on our SQL servers?

Any tips appreciated.

Thanks

Bill Baker