Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

NLB Cluster Intermittent SQL2K5 Connection Problems

3 views
Skip to first unread message

Mike Wazowski

unread,
May 27, 2009, 3:14:07 PM5/27/09
to
Hi Guys

I know this is really awkward but for the last three days now at around 5:30
PM our two (NLB clustered) webservers have been unable to connect to our SQL
2005 SP2 server. The SQL server installation is configured in a failover
cluster. The webservers are running Windows 2003 Standard Edition SP2 with
MDAC 2.8 SP2 (each has 4 quad core Xeons and 3 Gb RAM). The servers are
running classic ASP websites and both servers exhibit the common error
messages to the visitors:

DBNETLIB][ConnectionOpen (Connect()).]SQL Server does not exist or access
denied

DBNETLIB][ConnectionOpen (SECDoClientHandshake()).]SSL Security error

Rebooting both webservers resolves the problem. Another webserver running a
copy of the clustered websites, which is not used by the public, is still
able to access all the websites/databases that link to the SQL Server. We
don't think the problem is with the SQL Server set-up. When we failed over
from one to the other the problem persisted.

No changes or tasks have been made/added for months. Things have been
running smoothly for a while.

Nothing is logged in the event log on the webservers or the logs within SQL
Server. I don't know where to go next.

Does anyone have any advice?

Many thanks

Mike


Mike Wazowski

unread,
May 28, 2009, 3:37:38 PM5/28/09
to
Happened again tonight at 7:10 PM. This time I only rebooted one of the
cluster servers and it did the trick. It must be something to do with NLB
or sockets or something...

I don't know where to go. Nothing is logged anywhere.


"Mike Wazowski" <bookham_meas...@yahoo.com> wrote in message
news:%231bzQ9v...@TK2MSFTNGP03.phx.gbl...

Mike Wazowski

unread,
Jun 3, 2009, 9:04:02 AM6/3/09
to
Logged a call with Microsoft. After some Netstat analysis they have
suggested increasing the number of ports (MaxUserPorts) and the
TcpTimedWaitDelay registry settings on the two NLB servers. Apparently
they're nearing the default of 5000 so possibly out of ports.


"Mike Wazowski" <bookham_meas...@yahoo.com> wrote in message

news:%23v2tFv8...@TK2MSFTNGP05.phx.gbl...

Bob Barrows

unread,
Jun 3, 2009, 10:08:15 AM6/3/09
to
Good to know - thanks for the update.

It will be interesting to hear if this works for you. I've not run into
the problem myself, but I'm sure it could crop up someday.

--
HTH,
Bob Barrows


Mike Wazowski

unread,
Jun 8, 2009, 2:54:03 PM6/8/09
to
Yes, well it's Monday and we're past the previous times of failure so
fingers crossed. The two webservers in the NLB cluster have been putting
out between 35 and 40 Mbps each for most of the day so it's looking good so
far. Incoming is about 10Mbps but of course that's duplicated to both
servers.


"Bob Barrows" <reb0...@NOyahoo.SPAMcom> wrote in message
news:eDRT%23SF5J...@TK2MSFTNGP06.phx.gbl...

0 new messages