Hello all -

I've 'inherited' the sysadmin position for our company's online parts store. This is driven by the following:
2 Dell 2650's as Nodes 1 & 2; Dell 350 as Domain Controller for cluster; Powervault 220s Storage as Cluster and SQL Server resources (disk).
All 3 Dells run Windows 2000 which I am very careful to keep fully sync'd w/r/t to downloaded updates. SQL Server 2000 is the particular release I am running.

I've been administering this configuration for about 9 months now. It is an active / passive cluster and never gives me troubles, but I occassionally try the failover just to ensure everything is working.

This thing has never worked properly in my opion, as the failover frequently (90%+) results in our online parts store being unavailable. In every instance I can remember, the SQL Server comes up fine and I can run any kind of query against the database that I'd like to.

However, I find IIS (at least in a cluster) to be very, very unreliable.
In the past this has been due to account set up mis-configuration. After some stiff money to pay the company that originally setup the system to come in and re-install it, we (I thought) had the kinks worked out of the configuration and failover has been working fine.

As I've said, it hardly ever goes down, but when it does it is a real nightmare because IIS never works, which means the parts store is offline and management is asking me why we bought the cluster when it doesn't work (a fair question, but irritating for me to try and defend as I was not here when the decision to purchase a cluster was made. I can't defend the decision because I wouldn't bank my business on IIS ! But I digress...)

This weekend, I added 144GB (5 36GB Drives) of RAID 5 to the Storage Vault for backups. After getting the disks in, formatted and configured and brought online in the cluster, and added as disk resource to the SQL Server, I checked the failover once more.

Again as before, SQL Server works like a champ, but the parts store is offline when node 2 is goes active. This time IIS says it cannot find the server or has a DNS error! ! ! I know the netadmin has not changed the DNS configuration here, so again, for mysterious reasons, IIS has failed to do its job properly.

Does anyone else have these kinds of problems with IIS failovers in an SQL Cluster? If so, and you have some suggestions for troubleshooting, I'd be much obliged. MSFT has done a superb job on SQL Server, but IIS has left me very disillusioned and irritated. It would be nice if MSFT would put some effort into making the documentation more useful in troubleshooting and resolving problems. I'm personally hoping that Apache does a good job of supporting .NET (a strategic platform here), because if they do, I may put up an Apache server to run the parts store and forget about IIS !!!

<heavy sigh>

TiA for any help someone can lend. . .

I think that IIS works better in a Network Load Balancing
environment.

and SQL Server resources (disk).
2000 is the particular release I am running.
gives me troubles, but I occassionally try the failover
just to ensure everything is working.
store being unavailable. In every instance I can remember,
the SQL Server comes up fine and I can run any kind of
query against the database that I'd like to.
that originally setup the system to come in and re-install
it, we (I thought) had the kinks worked out of the
configuration and failover has been working fine.
means the parts store is offline and management is asking
me why we bought the cluster when it doesn't work (a fair
question, but irritating for me to try and defend as I was
not here when the decision to purchase a cluster was made.
I can't defend the decision because I wouldn't bank my
business on IIS ! But I digress...)
formatted and configured and brought online in the
cluster, and added as disk resource to the SQL Server, I
checked the failover once more.
time IIS says it cannot find the server or has a DNS
error! ! ! I know the netadmin has not changed the DNS
configuration here, so again, for mysterious reasons, IIS
has failed to do its job properly.
suggestions for troubleshooting, I'd be much obliged. MSFT
has done a superb job on SQL Server, but IIS has left me
very disillusioned and irritated. It would be nice if MSFT
would put some effort into making the documentation more
useful in troubleshooting and resolving problems. I'm
personally hoping that Apache does a good job of
supporting .NET (a strategic platform here), because if
they do, I may put up an Apache server to run the parts
store and forget about IIS !!!

Randy,

I've set up many clusters here in our lab using the IIS
failover without any problems, and I know this has also
been done at many of our customer sites out in the field.

Although the error message suggests a DNS lookup error I
suspect that this is not the case as your Domain
Controller/DNS server is external so I would expect the
same error from either node. (BTW I hope you have two
Domain Controllers/DNS servers as the cluster service does
very strange things when it cannot find a DC/DNS server)

It sounds to me as if the Web site on one of the nodes was
not set up with the same information as used on the first
node, or that the site was converted into a cluster
resource (which I suspect is more likely the case as you
do not mention that the IIS Instance Cluster Resource
fails)

Here's how I set my cluster IIS server up, which applies
for where the site information is on a local disk, or on a
shared disk. (our applications use both methods, and in
one case uses Tomcat to service the Java applets)

1. Use Internet Services Manager to set up a new web site
on NODE1. Set the IP address to be the IP address of the
Virtual Server that you wish to use - if the IP address is
not in the pull down list just type it in. (In theory you
should not use the same address as the SQL Server, but in
practice I have not found that to be a problem - if in
doubt create a new IP Address/Network Name pair and use
that) Set the home directory path to the path of your site
information. Stop this new web site once it is created.

2. If the site data is on a shared disk move the group
containing that disk to NODE2. (The disk should also be in
the same group as the IP address)

3. Create a new web site on NODE2 using exactly the same
parameters used on node one, including the description, IP
address and home directory.

4. Using Cluster Administrator create a new IIS Server
Instance resource in the same group. Both nodes are
possible onwers, and made it dependant on the IP Address
resource. (and shared disk if you like) At
the 'Parameters' window set the option 'WWW' and from
the 'IIS Server' pull down box select the name of the web
site that we created earlier.

5. Bring the new IIS Server Instance resource on-line,
take it offline, then bring it on-line again. (I've found
that this cycle gives the best results.

6. Check that you can conenct to the web site while the
group is one this node (NODE2).

7. Fail the group over to the other node and check the
operation there - depending on how your application
behaves your users 'may' have to log back in/reconnect to
the server - we have code in our application to
automatically reconnect.

Hope that helps
Kevin


and SQL Server resources (disk).
2000 is the particular release I am running.
gives me troubles, but I occassionally try the failover
just to ensure everything is working.
store being unavailable. In every instance I can remember,
the SQL Server comes up fine and I can run any kind of
query against the database that I'd like to.
that originally setup the system to come in and re-install
it, we (I thought) had the kinks worked out of the
configuration and failover has been working fine.
means the parts store is offline and management is asking
me why we bought the cluster when it doesn't work (a fair
question, but irritating for me to try and defend as I was
not here when the decision to purchase a cluster was made.
I can't defend the decision because I wouldn't bank my
business on IIS ! But I digress...)
formatted and configured and brought online in the
cluster, and added as disk resource to the SQL Server, I
checked the failover once more.
time IIS says it cannot find the server or has a DNS
error! ! ! I know the netadmin has not changed the DNS
configuration here, so again, for mysterious reasons, IIS
has failed to do its job properly.
suggestions for troubleshooting, I'd be much obliged. MSFT
has done a superb job on SQL Server, but IIS has left me
very disillusioned and irritated. It would be nice if MSFT
would put some effort into making the documentation more
useful in troubleshooting and resolving problems. I'm
personally hoping that Apache does a good job of
supporting .NET (a strategic platform here), because if
they do, I may put up an Apache server to run the parts
store and forget about IIS !!!

It sounds like a potential application problem. Just
incase you didn't know, during the failover, SQL Server
goes through a stop on the old node and a restart on the
new node. This means all connections into the DB will
need to reconnect.

If IIS or the application is persisting a connection to
the DB server, that's your problem. YOu need to account
for the connections being dropped and have a way to
reconnect.

Now, assuming you have IIS set up as another virtual
server (meaning it has its own group and such - and if
both SQL and IIS are on the same cluster, they should be
set up that way), it will fail independently of SQL
Server. The two are not linked. So if SQL Server fails,
it doesn't mean IIS will and vice versa.

NLB is a better choice for IIS, and you would need
separate HW for that. IIS and SQL work fine together, but
I do not know enough about your situation to diagnose.

Have you called MS Support yet?
and SQL Server resources (disk).
2000 is the particular release I am running.
gives me troubles, but I occassionally try the failover
just to ensure everything is working.
store being unavailable. In every instance I can remember,
the SQL Server comes up fine and I can run any kind of
query against the database that I'd like to.
that originally setup the system to come in and re-install
it, we (I thought) had the kinks worked out of the
configuration and failover has been working fine.
means the parts store is offline and management is asking
me why we bought the cluster when it doesn't work (a fair
question, but irritating for me to try and defend as I was
not here when the decision to purchase a cluster was made.
I can't defend the decision because I wouldn't bank my
business on IIS ! But I digress...)
formatted and configured and brought online in the
cluster, and added as disk resource to the SQL Server, I
checked the failover once more.
time IIS says it cannot find the server or has a DNS
error! ! ! I know the netadmin has not changed the DNS
configuration here, so again, for mysterious reasons, IIS
has failed to do its job properly.
suggestions for troubleshooting, I'd be much obliged. MSFT
has done a superb job on SQL Server, but IIS has left me
very disillusioned and irritated. It would be nice if MSFT
would put some effort into making the documentation more
useful in troubleshooting and resolving problems. I'm
personally hoping that Apache does a good job of
supporting .NET (a strategic platform here), because if
they do, I may put up an Apache server to run the parts
store and forget about IIS !!!

It all depends on what the application is doing. For
example, for stateless applications, SQL and IIS can go
together in the same failover cluster. However, I prefer
two independent clusters tiers (server farms) - one for
IIS (application stack) and the other for SQL server
(database stack). For stateful applications, I think
NLB/CLB or third-party load-balance clusters would do a
better job (of course SQL server failover cluster should
not be mixed).

The thumb rule would be - in a solution, it would be a
good practice to separate the application stack and
database stack in two distint tiers (for high-availability
of each function in respective cluster).

I welcome comments...

Hi Randy
We have had similar issues with a failover in IIS does not come up on the other node because of a DNS error
The solution we found was the DNS server specified was not dynamically updating the DNS resource name for our IIS cluster group. The DNS servers had a few issues.. One was a static entry in our DNS table (had to be removed), IPSEC had isolated the secondary node from the DNS server (hole addded in IPSec to allow dynamic updates to the closer faster DNS server), ... 2003 cluster allows you to specify wether to ignore the DNS lookup failure. I know there is a reg change in 2000 server that I found on MS technet site but do not remember the link.

It is not the cluster or IIS but a problem with how your DNS server is updating the reassignment of the cluster IP resource