Re: 4 Queries about DB clustering

From: Geoff N. Hiten (SRDBA_at_Careerbuilder.com)
Date: 02/11/05 

Date: Fri, 11 Feb 2005 14:34:57 -0500

You are getting closer. More comments inline. 

-- 
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com
I support the Professional Association for SQL Server
www.sqlpass.org
"Pankaj" <pankajchitriv@rediffmail.com(donotspam)> wrote in message
news:1D8FA145-7AF2-4EBE-9FA1-FD9822233F81@microsoft.com...
> Hi Geoff.
>
> Thankyou very much for your help. It is just great. I really really
> appreciate it.
>
> Further, to summarise our discussion what I understood to implement the
SQL
> Db Cluster is:
> 1. I need to have SCSI array, a Fiber array,  or a SAN as a shared device
> "on which I'll install SQL Server Database".
The database(s) will live on the shared storage.  SQL binaries will live on
the local disks of each cluster host.
>
> 2.  Say in a 2 node cluster, both the nodes would be typically Windows 2k3
> Server systems, with the SQL Server Instance reference (SQL Client) to the
> shared device on which SQL Server is installed.
A basic two-node SQL cluster will have four systems visible.  There are the
two physical host nodes.  There is the cluster vrtual server itself.  There
is the SQL Server virtual server.  Each virtual system must have a minimum
of one unique IP address, one unique network name, and a physical disk
device located on the shared array.  You access the SQL server by connecting
to the virtual server by name or IP address.
>
> 3. Say in Active / Passive mode, only 1 Win2k3 system will be utilized and
> in case of faliure, the other win2k3 server node will be active.
>
Correct.  The current terms are single-instance and multi-instance.  Each
new instance of SQL will have its own disk(s), IP address(es), and Network
name.  Once an instance is installed, you can set the preferred host order
to determine the "normal" home host for that instance.  You can have up to
16 instances on a single cluster.
> 4. I need to install SQL Client on both the nodes and Sql Server on shared
> device.
>
First you install MSCS (Clustering).  Windows 2003 has an excellent
clustering wizard that makes it very difficult to create a non-working
cluster.  Then you install a clustered instance of SQL server, running the
install CD from the node that currently owns the disk resource you wish to
use as the first SQL disk.  You can assign additional disks after
installation.  The installer writes the executable files to each host node
you choose as part of the installation.  This includes server and client
components.
> 5. Just wondering whether stored procedures of SQL DB will be executed on
> the  shared drive or on the Node. Does the shared drive have processing
> capabilities?
The binaries are on each host node.  The memory and CPU resources come from
the host node.  The data is stored on the shared array.  SQL uses a "Shared
Nothing" model so after installation you can run any instance from any host
node, regardless of whether any other node or instance is running.
>
> With this queries clarified , I'm ready to shoot :-).
>
> I really appreciate your help and request your comment on above queries.
>
> Regards,
> Pankaj A. Chitriv
>
>
>
>
>
> "Geoff N. Hiten" wrote:
>
> > Server clustering is available on Windows Server 2003 Enterprise
Edition.
> >
> > "Single Data Store" is not a service, it is a description.  All data
resides
> > on a single logical device physically connected to all host nodes.  This
can
> > be a SCSI array, a Fiber array,  or a SAN.  Ownership of this device is
> > arbitrated by the cluster service.  Microsoft SQL DB clustering exists
as a
> > failover technology only and has no distributed computing capabilities.
It
> > is not a scale-out technology.  One server owns the entire database and
no
> > other server can access the data without moving the entire SQL instance
to
> > another host.  Active-Passive and Active-Active are not accurate
> > descriptions of the current SQL clustering technology but are somewhat
> > misleading holdovers from an earlier version.
> >
> > Again, clustering is for availability and does not allow multiple
> > simultaneous connections to the data store from different host nodes.
> >
> > -- 
> > Geoff N. Hiten
> > Microsoft SQL Server MVP
> > Senior Database Administrator
> > Careerbuilder.com
> >
> > I support the Professional Association for SQL Server
> > www.sqlpass.org
> >
> > "Pankaj" <pankajchitriv@rediffmail.com(donotspam)> wrote in message
> > news:4A51BB01-D028-4B3B-9244-12EFA8469987@microsoft.com...
> > > Thanks Geoff.
> > >
> > > "Server Cluster" is a inbuild clustering service provided by Windows
2003
> > > Server.
> > >
> > > Just wondering, whether that Single Data Store is a seperate windows
> > system?
> > > If so what if that system breaks down? Is that the case that we do DB
> > > clustering only to distribute the processing capabilities?
> > >
> > > What I understood by Single Data Store means is there will be a system
on
> > > which SQL DB will be created and there would be multiple servers
accessing
> > > the same database. But this cannot be called as DB clustering as we
don't
> > > have to configure any VIP to access the single data store. Can you
clarify
> > > this please?
> > >
> > > Thanks a lot  again for your answers.
> > >
> > > Regards
> > > Pankaj A. Chitriv
> > >
> > > "Geoff N. Hiten" wrote:
> > >
> > > > Answers Inline
> > > >
> > > > -- 
> > > > Geoff N. Hiten
> > > > Microsoft SQL Server MVP
> > > > Senior Database Administrator
> > > > Careerbuilder.com
> > > >
> > > > I support the Professional Association for SQL Server
> > > > www.sqlpass.org
> > > >
> > > > "Pankaj" <pankajchitriv@rediffmail.com(donotspam)> wrote in message
> > > > news:847A2BE3-74E1-4B9E-9D61-EEE40D117155@microsoft.com...
> > > > > Hi, Below are few queries:
> > > > >
> > > > > 1. Say for 2 nodes cluster, Do we need to have 2 copies of
> > Databases( 1 on
> > > > > each node)
> > > > No.  Data is on shared storage and is controlled by one node at a
time.
> > > > >
> > > > > 2. Hows does the data Synchronization between different nodes of
> > databases
> > > > > takes place?
> > > > There is no synchronization.  There is a single data store.  Access
to
> > this
> > > > store is arbitrated by the cluster software.
> > > > >
> > > > > 3. In Active/Passive mode, does the passive node database is
exactly
> > in
> > > > the
> > > > > same state (in terms of data) as that of Active?
> > > > >
> > > > Clustering is failover not scaleout technology.
> > > >
> > > > > 4. Is 'Server Cluster' specifically for DB clustering ?
> > > > I am not sure what you mean by "Server Cluster"?
> > > > >
> > > > > Thanks in advance,
> > > > > Pankaj A. Chitriv
> > > > >
> > > >
> > > >
> > > >
> >
> >
> >

Relevant Pages

  • Re: Cluster IP Address change
    ... I've heard the same story from...after not hearing this in 2 years of SQL ... After changing sql ip address resource & sql name resource came online but ... sql server resource was failing. ... MVP - Windows Server - Clustering ...
    (microsoft.public.sqlserver.clustering)
  • Re: Side-by-side upgrade - moving master, msdb and model
    ... MVP - Windows Server - Clustering ... "In some cases, you may want to copy the system databases, including ... from the source SQL Server 2000 instance to the SQL ...
    (microsoft.public.sqlserver.clustering)
  • Re: Choosing correct high availability setup
    ... normal SQL startup, DB integrity check, roll back non-committed ... MVP - Windows Server - Clustering ... Basically there will be an Access database that arrives via FTP at ...
    (microsoft.public.sqlserver.clustering)
  • Re: How to install service pack 3a on a failover cluster
    ... MVP - Windows Server - Clustering ... We have two instances of sql running so I ... > the second node and rebooted the first node. ...
    (microsoft.public.sqlserver.clustering)
  • RE: Fulltext failure on a 2 node cluster
    ... Server full-text search resource online: "SQL Cluster Resource 'Full Text' ...
    (microsoft.public.sqlserver.clustering)
Quantcast