OK - Hi all
I have an irritating problem with a cluster and perhaps some of you can
short circuit it for me...

Environment
Windows Server 2003
Sql Server 2000 EE - SPK 3
3 node stretch cluster - only 3rd node is stretched

Problem Summary...
Node 1 (lets say 10.0.0.1) is ocassionally losing it ability to repond to
ping/name resolve. Trace route is failing one hop before reaching it.
However if we fail over to node 2 and back again all is ok for a random
period of time and then it fails to respond again
We are now leaving Node 2 as the Active node and witing to see if it happens
on node 2 also - if it does then I imagine the stakes will rise considerably
;-(

Things to Date...
All resources seem to be correctly configured and assigned to the correct
groups
The IP address is correct and unique for the SQL V Server resource
All n/w cards are functional
Nothing obvious in the event logs
The node responds to a ping locally (that is from Node1 over an RDP
session) without resorting to failover even though it is failing from an
external ping.

Other...
Not sure if this is related but at installation time I could not build the 3
nodes at once. I had to build Node 1 then add Node 2, then add Node 3.
Probably a red herring

Anyone ever have this or shed any light.
Should I be talking to the networking geniuses?

Thanks in advance

Barry Lynch

What are you using to stretch the cluster?

Cheers,

Rod

MVP - Windows Server - Clustering
http://www.nw-america.com - Clustering

"Barry Lynch" <allspammustdie (AT) yahoo (DOT) com> wrote

I saw this problem before and it was because a client had
the properties of the NIC set to update the DNS of the IP
information & the servers were behind a BIgIP (NAT). This
was hell on resolving and pings!

Good Luck, Rodney

Hi Rodney et al
That's very interesting - I will check tomorrow and let you know how I get
on
Thanks
Barry
"Rodney" <anonymous (AT) discussions (DOT) microsoft.com> wrote

To do a geographically dispersed cluster, you need to have
a solution that is in the Windows Catalog. How is the
third connected to a SAN? I assume it is mirrored ...

"Allan Hirt" <anonymous (AT) discussions (DOT) microsoft.com> wrote

Hi there Allan
The cluster is implemented as a VLAN over fibre so long as the ping
time to the stretched node is under 1500 ms (If I remember correctly
could be a differnt value - its 3am) then the cluster is viable - this
part works perfectly

With regard to this problem I'm beginning to work round to a mixture
of DNS and AD upgrade issues
Regards

Barry