Monday, March 12, 2012

Non sharing data cluster

Hi:

I'm trying to setup a MS Cluster but I don't know if it is feasible to
configure it in the way I think.

I have two machines with win2k server and MSSQL-2000 one of them is
currently performing as the production database and the other one is
the backup. The secondary one is keeping updated via the "Log
Shipping" technic.

We almost covered all the other possible failures of the other
components (ie: network, power, application servers, etc), the data is
maintaned in a raid which is ok but we want to cover also the
possibility of that failure too (yes, you can call us paranoids!! ).

The thing is we want to create a cluster that do not share the data,
but each machine of the cluster have thei own set of the same data.

My intention is to configure the cluster to detect a failure of one
machine and initiate the failover to the remaing one using the
appropiate scripts related to the promotion of a secondary server
keeped updateusing "log shipping".

Have anyone some experience with that kind of solution ??

Thanks in advance

Leonardo"Leonardo C" <lcomunia@.hotmail.com> wrote in message
news:fcc135d.0406150953.6fecf39a@.posting.google.co m...
> Hi:
> I'm trying to setup a MS Cluster but I don't know if it is feasible to
> configure it in the way I think.
> I have two machines with win2k server and MSSQL-2000 one of them is
> currently performing as the production database and the other one is
> the backup. The secondary one is keeping updated via the "Log
> Shipping" technic.
> We almost covered all the other possible failures of the other
> components (ie: network, power, application servers, etc), the data is
> maintaned in a raid which is ok but we want to cover also the
> possibility of that failure too (yes, you can call us paranoids!! ).
> The thing is we want to create a cluster that do not share the data,
> but each machine of the cluster have thei own set of the same data.
> My intention is to configure the cluster to detect a failure of one
> machine and initiate the failover to the remaing one using the
> appropiate scripts related to the promotion of a secondary server
> keeped updateusing "log shipping".
> Have anyone some experience with that kind of solution ??
> Thanks in advance
> Leonardo

I don't believe that's possible with a true cluster - in a cluster, if one
node goes down, its resources are immediately taken over by another node. A
cluster is intended for when you need absolute minimum downtime, but if you
can take the time to apply log shipping scripts, then I guess for you some
downtime is acceptable. In that case, why not just continue to use log
shipping?

But I must admit that I've never used clustering myself, so you might want
to post in microsoft.public.sqlserver.clustering to see if you can get some
more specific comments.

Simon|||Hi Simon:

Thanks for your response. Yes a minimum downtime is acceptable but as Log
Shipping needs a manual intervention a cluster can be the piece to automate
that part.
I'm forwarding this post to the Microsoft groups too, I'm not getting the
other one mentioned by you.

Maybe somebody knows how to automate this, without needing a Cluster.

"Simon Hayes" <sql@.hayes.ch> wrote in message
news:40cf3cc6$1_3@.news.bluewin.ch...
> "Leonardo C" <lcomunia@.hotmail.com> wrote in message
> news:fcc135d.0406150953.6fecf39a@.posting.google.co m...
> > Hi:
> > I'm trying to setup a MS Cluster but I don't know if it is feasible to
> > configure it in the way I think.
> > I have two machines with win2k server and MSSQL-2000 one of them is
> > currently performing as the production database and the other one is
> > the backup. The secondary one is keeping updated via the "Log
> > Shipping" technic.
> > We almost covered all the other possible failures of the other
> > components (ie: network, power, application servers, etc), the data is
> > maintaned in a raid which is ok but we want to cover also the
> > possibility of that failure too (yes, you can call us paranoids!! ).
> > The thing is we want to create a cluster that do not share the data,
> > but each machine of the cluster have thei own set of the same data.
> > My intention is to configure the cluster to detect a failure of one
> > machine and initiate the failover to the remaing one using the
> > appropiate scripts related to the promotion of a secondary server
> > keeped updateusing "log shipping".
> > Have anyone some experience with that kind of solution ??
> > Thanks in advance
> > Leonardo
> I don't believe that's possible with a true cluster - in a cluster, if one
> node goes down, its resources are immediately taken over by another node.
A
> cluster is intended for when you need absolute minimum downtime, but if
you
> can take the time to apply log shipping scripts, then I guess for you some
> downtime is acceptable. In that case, why not just continue to use log
> shipping?
> But I must admit that I've never used clustering myself, so you might want
> to post in microsoft.public.sqlserver.clustering to see if you can get
some
> more specific comments.
> Simon

No comments:

Post a Comment