Re: Cluster newbie: Majority Node Set

From: Mike Rosado [MSFT] (mikeros_at_online.microsoft.com)
Date: 02/21/05 

Date: Mon, 21 Feb 2005 16:57:41 -0600

Hi Gus,

I posted the following on an earlier thread. To answer your question
regarding Quorum and Majority Node Set (MNS), you would only want to use MNS
when setting up a Geographically Dispersed Clusters (GeoCluster) and even
then it has certain limitations when it comes to setting up software
application solutions.

Because a lot of software application solutions are based on Synchronous
Replication of data, so in order to have a GeoCluster using MNS you would
need a third party software that would handle Asynchronous Replication of
data.

Asynchronous Replication - Asynchronous replication means that if a change
is made to the data on site A, that change will eventually make it to site
B. Taking the same example as above, if an application at site A writes a
block of data to a disk mirrored to site B, then the IO operation will
complete as soon as the change is made to the disk at site A. The
replication software will transfer the change to site B in the background
and will eventually make that change to site B. Using asynchronous
replication; the data at site B may be out of date with respect to site A at
any point in time. Different vendors implement asynchronous replication in
different ways. Some preserve the order of operations, others do not. If a
solution preserves ordering, then the disk at site B may be out of date, but
it will always represent a state that existed at site A at some point in the
past. In other words, site B is crash consistent; the data at site B
represents the data at site A if site A had crashed at that point in time.
If a solution does not preserve ordering, the I/Os may be applied at site B
in an arbitrary order. In this case, the data set at site B may never have
existed at site A. Many applications can recover from crash consistent
states; very few (if any) can recover from out of order I/O sequences. In
short, never use asynchronous replication unless the order is preserved. If
order is not preserved, the data on site B may well appear corrupt to the
application and may be totally unusable.

Synchronous Replication - Synchronous replication means that if an
application performs an operation on one node at one site, then that
operation will not complete until the change has been made on the other
sites. Consider the case of synchronous, block level replication. If an
application at site A writes a block of data to a disk mirrored to site B,
then the IO operation will not complete until the change has been made to
the disk on site A and the disk on site B.

Microsoft recommends synchronous replication strategies because only they
can guarantee data consistency. For more information, please see the below
whitepaper on geographically dispersed clusters. Please keep in mind, that
if you decide to go this route, you should work with Microsoft and the
hardware vendor providing the solution to make sure that it can be
implemented in a supportable and non-performance impacting fashion.

Server Clusters: Majority Node Sets (a.k.a. Quorum of Nodes)
http://www.microsoft.com/technet/prodtechnol/windowsserver2003/technologies/clustering/majnode.mspx

Also if the hardware is not on the Windows Server Catalog under
Geographically Dispersed Cluster Solution, you may encounter other
difficulties. Because some hardware dependencies may be need for it too to
function correctly.
http://www.microsoft.com/windows/catalog/server/default.aspx?subID=22&xslt=category&pgn=904c28be-5a41-4db0-9c12-032dcb893c8b 

-- 
Hope this helps,
Mike Rosado
Windows 2000 MCSE + MCDBA
Microsoft Enterprise Platform Support
Windows NT/2000/2003 Cluster Technologies
====================================================
When responding to posts, please "Reply to Group" via your newsreader so
that others may learn and benefit from your issue.
====================================================
This posting is provided "AS IS" with no warranties, and confers no rights.
<http://www.microsoft.com/info/cpyright.htm>
-----Original Message-----
"Gus" <web_tur@yahoo.com> wrote in message
news:eTsKRaGGFHA.1292@TK2MSFTNGP10.phx.gbl...
> I have three servers running Windows 2000 Enterprise Edition. One of them
is
> the primary domain controller.
> We want to configure a cluster but we cannot afford hardware cluster
(shared
> disks, SAN, etc). It will be used for SQL Server Enterprise. We are
> considering log-shipping, transaction replication and other SQLServer
> methods but I think clustering is the best solution.
>
> So, searching MSDN I found the "Majority Node Set" model, but several
> question arised: (note: I understand that with 3 servers, only one can me
> down)
>
> 1) Data IS actually replicated in servers? I mean... one falls and the
> sqlserver data is still available for the other two servers.
>
> 2) What about this text that I found somewhere: "Windows Server 2003
> provides no mechanism to mirror or replicate user data across the nodes of
> an MNS cluster, so while it is possible to build clusters with no shared
> disks at all, it is an application specific issue to make the application
> data highly available and redundant across machines"
>
> 3) High availability must the hard-coded in the application? (I mean...
> connect to server1 and test, if not successfull connecto to server2 and
> test)
>
> 4) All three server will share the same IP?
>
> 5) Do we need extra hardware?
>
> This is all for now.
> Thanks for your time.
> Gus
>
>

Relevant Pages

  • Re: Server and exchange clustering over WAN
    ... With that said you don't need to cluster a DC anyway. ... Only if you have third party replication for the shared storage and two SANS. ... Exchange Server 2007 has other options when SP1 comes out, ... I've been tasked to set up a PDC/Exchange server that is currently running Windows server 2003 std R2 With a redundancy solution that would involve having another server at a remote site. ...
    (microsoft.public.windows.server.clustering)
  • Re: DFS Replication to a Cluster share target
    ... DFS Replication is not supported on shared storage of server clusters. ... to use the non-shared storage of a node of the cluster. ...
    (microsoft.public.windows.file_system)
  • Re: Cluster over Public WAN
    ... absolutely HAVE to set up a Private WAN? ... It will also require replication of the databases between the sites and software to provide not just the replication but locking so that the passive node does not have access to the disks in its local site. ... cluster to maintain consistancy and availablility. ... possible to use the same server as a node for another cluster (e.g SQL server ...
    (microsoft.public.windows.server.clustering)
  • Re: Cluster or Mirror
    ... Cluster or disk/volume replication level does not protect you ... against database/disk corruption. ... > automated switch to the standby server in case of a failure of the primary ...
    (microsoft.public.sqlserver.clustering)
  • Re: Cluster services - No shared disk
    ... You might want to look at "Majority Node Set" support in Microsoft Windows ... Majority node set server cluster configuration ... This posting is provided "AS IS" with no warranties, and confers no rights.Use of included script samples, if any, are subject to the terms specified at http://www.microsoft.com/info/cpyright.htm "JC-Unkwn" wrote in message ... >Can I use cluster services for just failover without shared disk? ...
    (microsoft.public.windows.server.clustering)
Quantcast