Choosing correct high availability setup

I have a client that wants me to start hosting their SQL and IIS for them.
They want to have the ability to have a "hot spare" in case the primary SQL
server fails. I have read and read until my eyes hurt about clustering and
mirroring and I am getting more confused on how to proceed.
My first question is, should I even be using IIS on the SQL box at all?
They have a web interface that gets it data from SQL.
My second question is, if I host their web site on a separate IIS box is
there any reason that I shouldn't go with database mirroring instead of any
other option?
Thanks for any help you can provide.
Marty
Never do both on the same box, read this -
http://msmvps.com/blogs/clusterhelp/archive/2006/02/17/84035.aspx.
Server Clustering with make the entire node HA, SQL mirroring will make that
database HA. Which does your customer require?
Cheers,
Rodney R. Fournier
MVP - Windows Server - Clustering
http://www.nw-america.com - Clustering Website
http://msmvps.com/clustering - Blog
http://www.clusterhelp.com - Cluster Training
ClusterHelp.com is a Microsoft Certified Gold Partner
"Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
message news:F03EC295-F9C8-4602-B787-BFBCC5841E48@.microsoft.com...
>I have a client that wants me to start hosting their SQL and IIS for them.
> They want to have the ability to have a "hot spare" in case the primary
> SQL
> server fails. I have read and read until my eyes hurt about clustering
> and
> mirroring and I am getting more confused on how to proceed.
> My first question is, should I even be using IIS on the SQL box at all?
> They have a web interface that gets it data from SQL.
> My second question is, if I host their web site on a separate IIS box is
> there any reason that I shouldn't go with database mirroring instead of
> any
> other option?
> Thanks for any help you can provide.
> Marty
|||Well that is what I had always thought, but you see more and more people
consolidating these functions.
Basically there will be an Access database that arrives via FTP at the web
server, the SQL server will pick it up from a shared drive and import it.
That will happen about every 15 minutes or so. The same web server will host
a site that the client can access to see real time data that it pulls from
the SQL server. The only snag is that if the primary SQL server goes down
for some reason, IIS will not know that it has to get that data from the
backup server if I am using only mirroring.
"Rodney R. Fournier [MVP]" wrote:

> Never do both on the same box, read this -
> http://msmvps.com/blogs/clusterhelp/archive/2006/02/17/84035.aspx.
> Server Clustering with make the entire node HA, SQL mirroring will make that
> database HA. Which does your customer require?
> Cheers,
> Rodney R. Fournier
> MVP - Windows Server - Clustering
> http://www.nw-america.com - Clustering Website
> http://msmvps.com/clustering - Blog
> http://www.clusterhelp.com - Cluster Training
> ClusterHelp.com is a Microsoft Certified Gold Partner
>
> "Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
> message news:F03EC295-F9C8-4602-B787-BFBCC5841E48@.microsoft.com...
>
>
|||Got it, use NLB for the IIS servers, Server Clustering for the backend
Cheers,
Rodney R. Fournier
MVP - Windows Server - Clustering
http://www.nw-america.com - Clustering Website
http://msmvps.com/clustering - Blog
http://www.clusterhelp.com - Cluster Training
ClusterHelp.com is a Microsoft Certified Gold Partner
"Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
message news:ECAE918E-4B1E-4339-A4CA-7F5752DC6768@.microsoft.com...[vbcol=seagreen]
> Well that is what I had always thought, but you see more and more people
> consolidating these functions.
> Basically there will be an Access database that arrives via FTP at the web
> server, the SQL server will pick it up from a shared drive and import it.
> That will happen about every 15 minutes or so. The same web server will
> host
> a site that the client can access to see real time data that it pulls from
> the SQL server. The only snag is that if the primary SQL server goes down
> for some reason, IIS will not know that it has to get that data from the
> backup server if I am using only mirroring.
>
> "Rodney R. Fournier [MVP]" wrote:
|||NLB?
If I use clustering for the backend, will the web server be able to keep
serving the data from the SQL database to the web interface in the case of a
disaster? The web site will be looking for a specific instance of SQL
correct?
"Rodney R. Fournier [MVP]" wrote:

> Got it, use NLB for the IIS servers, Server Clustering for the backend
> Cheers,
> Rodney R. Fournier
> MVP - Windows Server - Clustering
> http://www.nw-america.com - Clustering Website
> http://msmvps.com/clustering - Blog
> http://www.clusterhelp.com - Cluster Training
> ClusterHelp.com is a Microsoft Certified Gold Partner
>
> "Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
> message news:ECAE918E-4B1E-4339-A4CA-7F5752DC6768@.microsoft.com...
>
>
|||Yes NLB (Network Load Balancing), IIS is not made for Server Clustering, and
won't be supported/allowed with Windows Server 2008 Failover Clustering.
No matter how you do your SQL, IIS will need to know the instance name to
connect.
Cheers,
Rodney R. Fournier
MVP - Windows Server - Clustering
http://www.nw-america.com - Clustering Website
http://msmvps.com/clustering - Blog
http://www.clusterhelp.com - Cluster Training
ClusterHelp.com is a Microsoft Certified Gold Partner
"Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
message news:1429702D-83EF-4E83-B2DD-0C127B41DAC8@.microsoft.com...[vbcol=seagreen]
> NLB?
> If I use clustering for the backend, will the web server be able to keep
> serving the data from the SQL database to the web interface in the case of
> a
> disaster? The web site will be looking for a specific instance of SQL
> correct?
> "Rodney R. Fournier [MVP]" wrote:
|||Okay I will go with NLB for IIS. As far as the clustering for SQL goes, I am
very new on this subject. If one SQL box in the cluster fails, will the
other SQL box in that cluster pick up where the first left off and assume the
failed box's identity as far as IP address, DNS name, etc...?
Thanks for all of your help Rodney. I am glad there is a forum where people
like me can get real world help from people like yourself. Pat yourself on
the back man!!!
"Rodney R. Fournier [MVP]" wrote:

> Yes NLB (Network Load Balancing), IIS is not made for Server Clustering, and
> won't be supported/allowed with Windows Server 2008 Failover Clustering.
> No matter how you do your SQL, IIS will need to know the instance name to
> connect.
> Cheers,
> Rodney R. Fournier
> MVP - Windows Server - Clustering
> http://www.nw-america.com - Clustering Website
> http://msmvps.com/clustering - Blog
> http://www.clusterhelp.com - Cluster Training
> ClusterHelp.com is a Microsoft Certified Gold Partner
>
> "Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
> message news:1429702D-83EF-4E83-B2DD-0C127B41DAC8@.microsoft.com...
>
>
|||It will failover to another node and continue on, but at a cost. SQL not
running on the node until the failover, so the databases goes through the
normal SQL startup, DB integrity check, roll back non-committed
transactions, roll forward committed ones, etc. Your application(s) have to
be cluster aware to handle the failover. Ping tests will fail during the
process, though maybe only 4-5 depending on the network, hardware, DB sizes,
etc.
Cheers,
Rodney R. Fournier
MVP - Windows Server - Clustering
http://www.nw-america.com - Clustering Website
http://msmvps.com/clustering - Blog
http://www.clusterhelp.com - Cluster Training
ClusterHelp.com is a Microsoft Certified Gold Partner
"Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
message news:20D5F814-D38D-40B0-8BDE-2C50893F6509@.microsoft.com...[vbcol=seagreen]
> Okay I will go with NLB for IIS. As far as the clustering for SQL goes, I
> am
> very new on this subject. If one SQL box in the cluster fails, will the
> other SQL box in that cluster pick up where the first left off and assume
> the
> failed box's identity as far as IP address, DNS name, etc...?
> Thanks for all of your help Rodney. I am glad there is a forum where
> people
> like me can get real world help from people like yourself. Pat yourself
> on
> the back man!!!
> "Rodney R. Fournier [MVP]" wrote:
|||The only application that will depend on the SQL server would be IIS. How is
IIS going to react if one of the cluster nodes fail? I guess I would need to
point it to the alternate node's instance.
"Rodney R. Fournier [MVP]" wrote:

> It will failover to another node and continue on, but at a cost. SQL not
> running on the node until the failover, so the databases goes through the
> normal SQL startup, DB integrity check, roll back non-committed
> transactions, roll forward committed ones, etc. Your application(s) have to
> be cluster aware to handle the failover. Ping tests will fail during the
> process, though maybe only 4-5 depending on the network, hardware, DB sizes,
> etc.
> Cheers,
> Rodney R. Fournier
> MVP - Windows Server - Clustering
> http://www.nw-america.com - Clustering Website
> http://msmvps.com/clustering - Blog
> http://www.clusterhelp.com - Cluster Training
> ClusterHelp.com is a Microsoft Certified Gold Partner
>
> "Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
> message news:20D5F814-D38D-40B0-8BDE-2C50893F6509@.microsoft.com...
>
>
|||Correct as Geoff "the SQL God and good buddy of mine" already stated.
Cheers,
Rodney R. Fournier
MVP - Windows Server - Clustering
http://www.nw-america.com - Clustering Website
http://msmvps.com/clustering - Blog
http://www.clusterhelp.com - Cluster Training
ClusterHelp.com is a Microsoft Certified Gold Partner
"Marty Shifflett" <MartyShifflett@.discussions.microsoft.com> wrote in
message news:7871ABB2-D689-41A6-B9F7-30C5482A4EE3@.microsoft.com...[vbcol=seagreen]
> So the node that picks up truly is a "clone" of the failed one? Well then
> it
> sounds like I may be better off going with clustering than mirroring in my
> situation wouldn't you say?
> "Rodney R. Fournier [MVP]" wrote: