RE: Windows 2008 failover cluster question

The only option other than shared storage is to add a third party application
to provide replciation between the local disk on two nodes. GoeCluster does
this. I am sure there are others.

This is something that, IMHO, Microsoft needs to address. Shared storage is
a huge single point of failure and customers should have other options.

"ggcluster" wrote:

>
> Hi guys,
>
> I need to setup a test environment and configure failover cluster on
> win 2008. I have 2 servers with local storage, ide channels, and
> ovbviously this doesnt work.
>
> If I was to buy 2 servers with locally attached SCSI disks, will that
> work, or does it HAVE to be shared storage?
>
>
> --
> ggcluster
> ------------------------------------------------------------------------
> ggcluster's Profile: http://forums.techarena.in/members/72933.htm
> View this thread: http://forums.techarena.in/server-cluster/1218232.htm
>
> http://forums.techarena.in
>
>

"cojits" <> wrote in message
news:A692DF80-C8F0-4581-877C-...
> The only option other than shared storage is to add a third party
> application
> to provide replciation between the local disk on two nodes. GoeCluster
> does
> this. I am sure there are others.
>
> This is something that, IMHO, Microsoft needs to address. Shared storage
> is
> a huge single point of failure and customers should have other options.

Do you mean by doing something like making Windows Storage Server into an
iSCSI target?

Do you mean like what the Exchange team has done with CCR in 2007 and with
the Database Availability Group technology?

Are you asking Microsoft to build a multi-site replication technology with
disk locking mechanisms to support multi-site clustering for all
applications?

I am just trying to figure out what everyone wants. <G>

On a serious note, site resiliency is becoming more and more of a focus for
Microsoft with all of its products. To me, there are several issues with
site resiliency:
1. How are we going to redirect clients to the new location in the event of
a site disaster?
2. What will out time requirements be in the event of a disaster?
3. How we will respond to the disaster, will it be automated, or will some
intervention be needed to protect against unwanted failovers?
4. What supporting services are required for which applications and how will
they be managed (i.e. AD sites and AD replication as well as DNS)?

High availability is vital for many businesses, without a doubt. But which
apps, and what SLAs are really needed? Most importantly, what components of
HA are needed, meaning do we need server redundancy, application/service
availability, and/or site resiliency?

Just my thoughts on the matter...

--
Russ Kaufmann,
MVP, MCSE: Messaging and Security, MCT, MCITP, MCTS and other stuff

ClusterHelp.com, a Microsoft Certified Gold Partner
Web http://www.clusterhelp.com
Blog http://msmvps.com/clusterhelp

Whoa there! I think I touched a nerve!

Russ, my point was that the idea of using a single disk array in one
location for all of your server operations is short-sighted, not that
Microsoft is responsible for that idea. Sort of like those date-coding
shortcuts that every software developer used twenty-five years ago...

Every buisness has to decide how much resiliency is required, but going for
site resiliency also takes care of high availability. I'm working on the
redirect issue right now. For some applications and services, it's just not
that big of a deal. Software devs are going to have to stop this nonsense of
hard coding IP addresses, though!

HA, redundancy and site resiliency is something that all of IT is going to
have to think about. Clustering is great, but unless you can cluster to
another physical location, you still leave yoruself open to any kind of event
that comprimises that server room.

"Russ Kaufmann" wrote:

> "cojits" <> wrote in message
> news:A692DF80-C8F0-4581-877C-...
> > The only option other than shared storage is to add a third party
> > application
> > to provide replciation between the local disk on two nodes. GoeCluster
> > does
> > this. I am sure there are others.
> >
> > This is something that, IMHO, Microsoft needs to address. Shared storage
> > is
> > a huge single point of failure and customers should have other options.
>
> Do you mean by doing something like making Windows Storage Server into an
> iSCSI target?
>
> Do you mean like what the Exchange team has done with CCR in 2007 and with
> the Database Availability Group technology?
>
> Are you asking Microsoft to build a multi-site replication technology with
> disk locking mechanisms to support multi-site clustering for all
> applications?
>
> I am just trying to figure out what everyone wants. <G>
>
> On a serious note, site resiliency is becoming more and more of a focus for
> Microsoft with all of its products. To me, there are several issues with
> site resiliency:
> 1. How are we going to redirect clients to the new location in the event of
> a site disaster?
> 2. What will out time requirements be in the event of a disaster?
> 3. How we will respond to the disaster, will it be automated, or will some
> intervention be needed to protect against unwanted failovers?
> 4. What supporting services are required for which applications and how will
> they be managed (i.e. AD sites and AD replication as well as DNS)?
>
> High availability is vital for many businesses, without a doubt. But which
> apps, and what SLAs are really needed? Most importantly, what components of
> HA are needed, meaning do we need server redundancy, application/service
> availability, and/or site resiliency?
>
> Just my thoughts on the matter...
>
> --
> Russ Kaufmann,
> MVP, MCSE: Messaging and Security, MCT, MCITP, MCTS and other stuff
>
> ClusterHelp.com, a Microsoft Certified Gold Partner
> Web http://www.clusterhelp.com
> Blog http://msmvps.com/clusterhelp
>

Ack! My display name change happened mid-thread...sorry about that.




I'm just saying that relying on a single disk array stored in the same rack
as all of your servers isn't exactly a long-term solution. SOoner or later
something is goign to happen to that site.

"Jeff Strubberg" <> wrote in message
news:CCA345EB-4339-4449-B65B-...
> Whoa there! I think I touched a nerve!

No nerver, I was just trying to figure out what you were requesting from
Microsoft vs. what has already been provided in one shape or another.

I have a very thick skin. <G>

> Russ, my point was that the idea of using a single disk array in one
> location for all of your server operations is short-sighted

My point would be, we have the ability, in many cases, to create copies of
the data in a remote location already depending upon the
application/service.

> Every buisness has to decide how much resiliency is required, but going
> for
> site resiliency also takes care of high availability.

I would disagree. For example, while Microsoft might refer to SCR in
Exchange Server 2007 as HA, I do not since it requires manual intervention
and is not an automated process.


> For some applications and services, it's just not
> that big of a deal. Software devs are going to have to stop this nonsense
> of
> hard coding IP addresses, though!

Agreed.

> HA, redundancy and site resiliency is something that all of IT is going to
> have to think about. Clustering is great, but unless you can cluster to
> another physical location, you still leave yoruself open to any kind of
> event
> that comprimises that server room.

This is open to discussion. Again, going back to the SCR in Exchange Server
2007 feature, it provides site resiliency, but it is not automated, and this
may be a good thing for some as it does require manual intervention with a
decision to implement based on circumstances. For example, if power is lost
to the datacenter but it will return within 20 minutes, you would not want
to failover to another location and force all clients to connect across a
WAN link.

I contend that site resiliency should not be something that is automated in
most cases. Yet, back to the discussion, the option exists for many
applications and services without using multi-site failover clusters.

--
Russ Kaufmann,
MVP, MCSE: Messaging and Security, MCT, MCITP, MCTS and other stuff

ClusterHelp.com, a Microsoft Certified Gold Partner
Web http://www.clusterhelp.com
Blog http://msmvps.com/clusterhelp

"Jeff Strubberg" <> wrote in message
news:60FF668A-A3CF-43C1-8C1F-...
> Ack! My display name change happened mid-thread...sorry about that.
>
>
>

Quote:

> Are you asking Microsoft to build a multi-site replication > technology with >> disk locking mechanisms to support multi-site clustering for all >> applications?

>
> I'm just saying that relying on a single disk array stored in the same
> rack
> as all of your servers isn't exactly a long-term solution. SOoner or
> later
> something is goign to happen to that site.

Sure. Something will eventually happen, but do you want to spend money based
upon your facility being hit by a meteor? How about a 100 year flood level?
How about for a potential tornado strike?

What I do in consulting engagements is help companies define all of the
potential situations that will result in an application or service failure
and then figure which ones require mitigation and which ones just are not
worth developing mitigation plans.

Most organizations that develop HA plans do not put each node and the
storage in the same rack. That would be poor planning, but it does happen.
Most organizations that plan properly will have:

- Nodes in different racks
- Nodes on different power arrays
- Nodes on different network switches
- Nodes in different rooms within the datacenter
- Nodes using different secondary power and cooling providers
- Nodes using different network gateways using different external ISPs
- and so on...

The biggest issue is whether site resiliency is needed and then making the
decision as to whether it needs to be an automated failover solution or a
solution that requires management decision making for failover and then a
manual process that is followed.


--
Russ Kaufmann,
MVP, MCSE: Messaging and Security, MCT, MCITP, MCTS and other stuff

ClusterHelp.com, a Microsoft Certified Gold Partner
Web http://www.clusterhelp.com
Blog http://msmvps.com/clusterhelp