[Linux-cluster] Queries regarding A four node Cluster setup

[Sachin V Renake]

Hello Team
Need your help in following cluster setup.

---- There are four server in the cluster like server A, server B,server C 
and server D.
---- Server A and server B is installed with operating system Redhat linux 
As 3.0 and Beblogic application
---- Server C is installed with operating system Redhat linux As 3.0 and 
Oracle database.
--- Server D is failover server which is installed with operating system 
Redhat linux As 3.0 and Beblogical application and Oracle database.
--- Lvm's are created in the Hp storage box which will be the filesystems 
for the application servers and database server and they are mapped to the 
particular servers.
--- Server D is mapped with all the filessystems created on HP storage box 
but they are not mounted in this server.
--- Filesystem which has been created in HP storage box is a pure data 
which is there for application server and database servers.
--- So Server A  and server B are communicating with server C which is the 
Databse (Oracle) server.
--- Now the scenario will be like this from End user side.People are going 
to logg in Server A or Server B or Server C as per their requirement and 
they are going to read and write their respective data related to their 
functions on the  storage Box.
--- Clumanager is cluster management program which is to be used.
--- Filesystem in 

So now ,customers ONLY requirement is that in case Server A is going down 
because of some reason,people (End users) still should be in position to 
continue with their Data I/O on filesystem which is lieing on HP storage 
box so our cluster will function is such a way that the particular 
filesystem which was mounted in Server A which now will be transfered to 
Server D so i/o will not hamper.
One point i want to make clear that during such instance,as per confirmed 
with customer,the partcular  Beblogic application will be started by human 
intervention i.e manually in Server D.

Now our queries are like this:

1. We have found that as the filesystem which has been mapped to server A 
is also mapped to server D and can be mounted simultaneously in both the 
servers and can be accessed and we can carry out I/O on it.So this can 
lead to corruption of data and causes insecurities.Such thing  is not 
going to be carry out by anyone but accidently can be happened. Suppose if 
server A hangs for small period of time and though cluster manager 
software,the filesystem which was assigned to this server will 
automatically get mounted in Failover server D .i.e Service(Resources) 
will get transfered.And if Server A is alive again after that period so 
Are there any possibilities that Both servers can mount and access the 
Same filesystem and cause Data corruption?
Please elaborate the functionality of Watchdog utility in this case.

2. If suppose in future,if we need to add extra Luns to the existing 
server which is part of cluster,from Hp storage box,so general practice is 
to use commands like
rmmod qla2300---------------------------------in order to remove the 
module of hba card
modproble qla2300-------------------------- in order to install module 
again
But command will involve the removal of device driver means the Detachment 
(Not physical) of the externally connected storage so such situtation will 
be like Downtime of the server.
So query is that is there any fascility in Redhat so that Addition of new 
Lun will not involve removal of the driver Module and detachment(Not 
physical)of the storage so we can carry out this activity on the Fly 
without disturbing the setup?

Warm Regards
Sachin Vishwanath Renake
IBM Global Services Pvt.Ltd.
The IL & FS Building,
4th Floor, "A" wing,
Bandra Kurla Complex,Bandra (E),
Mumbai 400 051.
Telephone# 56962020 Board
Ext# 2202
Email : sachin.renake at in.ibm.com

"Don't judge each day by the harverst you reap,but by the seeds you plant" 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20050610/fe51adf0/attachment.htm>