Thanks a lot guys. For your suggestion. it really helped me a lot 

MOHAMMED ASHRAF

MOB: 9870161983

--- On Wed, 10/3/10, Nadeem M. Khan <[email protected]> wrote:

From: Nadeem M. Khan <[email protected]>
Subject: Re: [LinuxVadaPav] help on cluster server
To: [email protected]
Date: Wednesday, 10 March, 2010, 8:04 PM







 



  


    
      
      
      On Wed, Mar 10, 2010 at 2:21 PM, ashraf mohammed <ashraflinux@ yahoo.com> 
wrote:

>

>

>

> HI guys

> i want to know wat is cluster server on linux .And how to configure it. and 
> can i do clustering of server on my home pc.i am using vmware 6.0.5. i can 
> add as many server as a want in vmware. So can i learn clustering at home.

> i want to learn basic clustering first. so guys can u help out with some 
> ideas.

> how to achieve this...

> or any documentation which will help me in better way



Hi Ashraf,



I wouldn't get too much into the theory of clusters, but want to share

info from a practical point of view. Clustering is used to accomplish

either of two objectives - High Availability or Load Balancing. There

are tons of docs for the latter so I'll leave it for you to read those

up.



As far as HA is concerned, the objective if to keep your resource

(application or database or whatever) you want always available. The

operation is like so: You have two nodes running Linux or Unix and a

shared storage (either fibre channel or nfs, but fc has better

performance) . For example, lets consider you have an apache webserver

that you want to be highly available. For this, you would have a

volume group residing on the SAN and which is shared across both

nodes. This volume group will contain apache's document root and other

files required for apache's operation. Each node will have at least

two IP addresses. One service IP used for accessing the webserver and

another private IP used to connect the two nodes for heart beat. These

IPs are in different subnets.Let' s say your primary node is NodeA and

secondary is NodeB. Under normal operation, when uses access your

website, they will be directed to NodeA and everything will be fine.

The cluster software monitors the availability of each node using the

heatbeat IPs. Lets say NodeA crashes, the cluster softare will miss a

few heatbeats and detect that NodeA has crashed. The service IP will

then move to NodeB. Teh VG will be moved to NodeB and varied on in

exclusive mode. This is termed as failover. When users access your

webserver now, they will be directed to NodeB. All this is transparent

to the users and they will not be aware of the failover except for the

time it actually takes for the failover to complete.



When you fix NodeB and bring it back up, the cluster heartbeat will

detect this and move the service IP back to NodeA. The VG will be

varied off on NodeB and move to NodeA. This is termed as failback.

Users will be directed to NodeA now as things are back to normal.



Some examples of HA clustering software:



Linux: RHCS, Veritas, Linux-HA

AIX: HACMP

HP-UX: ServiceGuard

Solaris: Veritas.

Oracle DB: RAC. RAC is a bit different as it uses the concept of

virtual IP (VIP) instead of service IP.



This was just a simple example involving IP and VG. You basically have

scripts to determine what action needs to be taken during failover and

failback.



It is possible to install a Linux cluster on Vmware. I remember

reading a doc online but for the life of me can't recollect where.

Google is your friend.



Regards,

NMK.



    
     

    
    


 



  






      Your Mail works best with the New Yahoo Optimized IE8. Get it NOW! 
http://downloads.yahoo.com/in/internetexplorer/

[Non-text portions of this message have been removed]

Reply via email to