Groups | Blog | Home
all groups > sql server clustering > january 2004 >

sql server clustering : Mixed Windows 2000/2003 in cluster problem


pdube NO[at]SPAM clic.net
1/14/2004 8:13:49 AM
For 3 years I had a Windows 2000 server farm consisting of 4 servers
(2 web servers load balanced and 2 sql servers for clustering). The 4
servers are in domain. They are the only servers in that domain.

I had a hard disk crash on 1 SQL server recently, so I decided to try
reinstalling it with W2003.

So I install a new server with Windows 2003, integrated it in Active
Directory, add the node to the SQL cluster, re-installed the virtual
SQL server to the new node, applied all patches (Windows 2003 and SQL
2000 SP3a).

The cluster seems to work fine: it fails over nicely.
But from my web servers I am unable to communicate with my virtual SQL
server when it runs on the 2003 box. When it runs on the 2000 box,
everything is fine, but if I fail over on the 2003 box I get errors
like "Specified SQL server not found. ConnectionOpen (Connect())".

All network functionnalities seems to be very OK. From my workstation
(which is not in the same domain, but in a trusted domain), I can
communicate with the database (whether it runs on the 2000 or 2003
box).
The web server pings the 2003 SQL server.

Installing all boxes with 2003 is not an option at this point. I need
to be able to do this in steps.

Paul Dube
1/14/2004 2:42:04 PM
Now, for god-knows-what-reason, I am unable to fail over to the 2003
box.

I get these errors in Event Viewer

Event ID: 17052
Description: [sqsrvres] ODBC sqldriverconnect failed

then

Event ID: 17052
Description: [sqsrvres] checkODBCConnectError: sqlstate = 08001;
native error = 6; message = [Microsoft][ODBC SQL Server Driver][TCP/IP
Sockets] Specified SQL server not found.

I closely followed the required network configuration at
http://support.microsoft.com/?id=258750
but it does not help.

What could be wrong in my configuration???

Paul Dube
1/14/2004 2:48:43 PM
Just to add some more information:

When I say the SQL does not fail over, it's the SQL Server that does
not fail over. I mean the first 3 Resources do fail over (Physical
Disk, IP Address and Network Name), but the remaining 3 resources get
stuck...

Geoff N. Hiten
1/14/2004 4:13:23 PM
Apply Service Pack 3a to your entire SQL instance.

--
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com


[quoted text, click to view]

Paul Dube
1/14/2004 4:44:46 PM
All instances already run SP3a. All patches applied everywhere.

Paul

On Wed, 14 Jan 2004 16:13:23 -0500, "Geoff N. Hiten"
[quoted text, click to view]
Geoff N. Hiten
1/14/2004 10:10:58 PM
It looks line the SQL Server dependancies come online just fine, but the
server app itself doesn't. See if the Application event log shows why.
Also, just for grins, rerun the SP3a install from the 2003 node with the
cluster resource running on the 2000 node. That will make sure the binaries
are up to date on the 2003 node without affecting the running system.

--
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com





[quoted text, click to view]

Paul Dube
1/15/2004 11:05:13 AM
Done. It looks like it has trouble doing the remote setup. I guess it
should not matter too much since it was done locally on that server.

The behavior is still the same: unable to fail over.

Let's call my servers SQL1 (Windows 2003) and SQL2 (Windows 2000).

On SQL1, in my Application Log: I see this when I try to fail from
SQL2 to SQL1:

Event ID: 17052
Description: [sqsrvres] ODBC sqldriverconnect failed

then

Event ID: 17052
Description: [sqsrvres] checkODBCConnectError: sqlstate = 08001;
native error = 6; message = [Microsoft][ODBC SQL Server Driver][TCP/IP
Sockets] Specified SQL server not found.

then

Event ID: 17052
Description: [sqsrvres] ODBC sqldriverconnect failed

then

Event ID: 17052
Description: [sqsrvres] checkODBCConnectError: sqlstate = 01000;
native error = 2afc; message = [Microsoft][ODBC SQL Server
Driver][TCP/IP Sockets] Specified SQL server not found.

then some more of the same errors when retrying before it aborts and
revert back to SQL2.

What do you suggest? These are test servers. I can do almost anything
I want on them, but I need to be able to have an efficient upgrade
plan when I'll do this on my production servers (that is with very few
downtime).

Thanks!

On Wed, 14 Jan 2004 22:10:58 -0500, "Geoff N. Hiten"
[quoted text, click to view]
Geoff N. Hiten
1/15/2004 12:02:17 PM
I think you should drop back to square one. Take SQL1 out of the SQL system
and reinstall it completely.

Follow the instructions in BOL precisely for removing a node from SQL.
Remove node SQL1.

I would reboot both nodes, just to clear everything. It looks like the
resources are correct for node SQL1 so the cluster service is likely OK.
Make sure the SQL service account is a domain account with local admin privs
on SQL1.

Again following BOL exactly, follow the instructions to add a node to a
virtual server. All of these topics are under 'Maintaining a Failover
Cluster'. Note that all of the configuration of the virtual server on the
'new' node is done by SQL install.

Reboot SQL1 and apply Service Pack 3a to SQL1 Only. Run the install from
the console of SQL1 and the service pack will update local binaries only.

Also, make sure the account you are running the install from is a domain
account with local admin privs on both boxes. Make sure you have logged
into the console of SQL1 with this account at least once before running the
install. Other wise some temporary directories that pre-install wants will
not be there.

--
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com




Rerun the SQL installation on Node 2, advanced options, maintain
[quoted text, click to view]

cgross NO[at]SPAM online.microsoft.com
1/30/2004 4:29:00 PM
Check your aliases (cliconfg.exe) also. " Specified SQL server not found"
could indicate something weird in an alias.

Cindy Gross, MCDBA, MCSE
http://cindygross.tripod.com
This posting is provided "AS IS" with no warranties, and confers no rights.
AddThis Social Bookmark Button