Clustering with 2 active nodes and 1 passive node

Normally, clustering is setup as a 1-to-1 relationship between active and
passive nodes, at least that is my understanding. We have a situation where
we are being asked to setup a 2 active node, 1 passive node environment.
Here's the question - if one of the active nodes fails, it fails over to the
passive node. Now what happens if the 2nd active node fails? Will the other
active node (originally the passive node) handle the failover or will it try
to failover to the first active node that failed or will the whole thing just
grind to a screeching halt? I know we can setup a 2 active/ 2passive node
setup, but we wanted to research the 2-to-1 possibility as well.
Thanks!
Message posted via http://www.droptable.com
Hi
You don't need a 1:1 relationship between nodes.
And forget the terms active / passive, it is totally out of date and does
not describe SQL Server 2000/2005 clustering.
You have an instance of SQL Server running on an node.
If you have 2 nodes, it could run on any of those nodes at any one time. If
you had 3 nodes, it could run on any of those nodes, at any one time.
If you have 2 nodes, one sits around waiting for a failover. If you have 3
nodes, 2 sit around waiting for a failover.
If you have more than one SQL Server instance running on the cluster, then
the instances can run on any of the nodes.
The instances could be all running on the same node, or dispersed over the
various nodes.
Quite often, people run with 3 nodes and 2 instances or 4 nodes with 3
instances. There is nothing stopping you running 3 instances on 3 nodes, 4
instances on 4 nodes, or any combination of up to 16 instances over 4 nodes.
You just need to make sure that you have enough resources, that in worst
case, one node could run all instances.
Regards
Mike Epprecht, Microsoft SQL Server MVP
Zurich, Switzerland
IM: mike@.epprecht.net
MVP Program: http://www.microsoft.com/mvp
Blog: http://www.msmvps.com/epprecht/
"Michael G via droptable.com" <forum@.droptable.com> wrote in message
news:541CE6C54F90F@.droptable.com...
> Normally, clustering is setup as a 1-to-1 relationship between active and
> passive nodes, at least that is my understanding. We have a situation
> where
> we are being asked to setup a 2 active node, 1 passive node environment.
> Here's the question - if one of the active nodes fails, it fails over to
> the
> passive node. Now what happens if the 2nd active node fails? Will the
> other
> active node (originally the passive node) handle the failover or will it
> try
> to failover to the first active node that failed or will the whole thing
> just
> grind to a screeching halt? I know we can setup a 2 active/ 2passive node
> setup, but we wanted to research the 2-to-1 possibility as well.
> Thanks!
>
> --
> Message posted via http://www.droptable.com
|||Very helpful. Question: So now, even with RAID 10, the single point of
failure becomes the shared drive array? How does one go about creating a
failover for this?
Respectfully,
Michael Wiederhold
michael@.auctionarms.com
"Mike Epprecht (SQL MVP)" <mike@.epprecht.net> wrote in message
news:%230j6BCZtFHA.3236@.TK2MSFTNGP14.phx.gbl...
> Hi
> You don't need a 1:1 relationship between nodes.
> And forget the terms active / passive, it is totally out of date and does
> not describe SQL Server 2000/2005 clustering.
> You have an instance of SQL Server running on an node.
> If you have 2 nodes, it could run on any of those nodes at any one time.
> If you had 3 nodes, it could run on any of those nodes, at any one time.
> If you have 2 nodes, one sits around waiting for a failover. If you have 3
> nodes, 2 sit around waiting for a failover.
> If you have more than one SQL Server instance running on the cluster, then
> the instances can run on any of the nodes.
> The instances could be all running on the same node, or dispersed over the
> various nodes.
> Quite often, people run with 3 nodes and 2 instances or 4 nodes with 3
> instances. There is nothing stopping you running 3 instances on 3 nodes, 4
> instances on 4 nodes, or any combination of up to 16 instances over 4
> nodes. You just need to make sure that you have enough resources, that in
> worst case, one node could run all instances.
> Regards
> --
> Mike Epprecht, Microsoft SQL Server MVP
> Zurich, Switzerland
> IM: mike@.epprecht.net
> MVP Program: http://www.microsoft.com/mvp
> Blog: http://www.msmvps.com/epprecht/
> "Michael G via droptable.com" <forum@.droptable.com> wrote in message
> news:541CE6C54F90F@.droptable.com...
>
|||Hi
Most high end SAN's have a block level copy mechanism so that you can
"mirror" that data to another SAN, preferably at another location.
EMC has SRDF that can run synchronous or asynchronous over fast WAN links.
Tape backups must never be forgotten either.
Regards
Mike Epprecht, Microsoft SQL Server MVP
Zurich, Switzerland
IM: mike@.epprecht.net
MVP Program: http://www.microsoft.com/mvp
Blog: http://www.msmvps.com/epprecht/
"michael wiederhold" <michael@.auctionarms.com> wrote in message
news:%23RhSzuztFHA.3068@.TK2MSFTNGP14.phx.gbl...
> Very helpful. Question: So now, even with RAID 10, the single point of
> failure becomes the shared drive array? How does one go about creating a
> failover for this?
> Respectfully,
> Michael Wiederhold
> michael@.auctionarms.com
> "Mike Epprecht (SQL MVP)" <mike@.epprecht.net> wrote in message
> news:%230j6BCZtFHA.3236@.TK2MSFTNGP14.phx.gbl...
>
|||Thanks. Really do appreciate your participation in this newsgroup.
Respectfully,
Michael Wiederhold
michael@.auctionarms.com
"Mike Epprecht (SQL MVP)" <mike@.epprecht.net> wrote in message
news:ONHGyz8tFHA.2540@.TK2MSFTNGP09.phx.gbl...
> Hi
> Most high end SAN's have a block level copy mechanism so that you can
> "mirror" that data to another SAN, preferably at another location.
> EMC has SRDF that can run synchronous or asynchronous over fast WAN links.
> Tape backups must never be forgotten either.
> Regards
> --
> Mike Epprecht, Microsoft SQL Server MVP
> Zurich, Switzerland
> IM: mike@.epprecht.net
> MVP Program: http://www.microsoft.com/mvp
> Blog: http://www.msmvps.com/epprecht/
> "michael wiederhold" <michael@.auctionarms.com> wrote in message
> news:%23RhSzuztFHA.3068@.TK2MSFTNGP14.phx.gbl...
>