[Beowulf] HA OSCAR for loadbalancing and failover

Michael Will mwill at penguincomputing.com
Fri Mar 4 15:52:18 PST 2005


Even though it is an interesting idea to use a beowulf cluster for this,
in particular when using several nodes to do loadbalancing with and
automatic deployment of services, I think it is the wrong tool for the
task you have set for yourself.

Your requirements would probably be more easily fulfilled with a simple
HA failover cluster (no oscar involved). See http://www.ultramonkey.org for details.

Especially when you only have two servers, one as primary and one as standby, 
which is a classical active/passive config, then there is no reason to have the 
complexity of a beowulf style cluster. 

ultramonkey.org also mentions LVS which helps with loadbalancing and I believe
they even have a solution for session synchronisation, which means that even
when a failover of the loadbalancer occurs, a tcp/ip session does not die but 
gets redirected.

You will not need any PBS then, but rather have a package called 'heartbeat' that
defines the services to be failed over in it's own config files.

Michael

On Thursday 03 March 2005 10:51 pm, Rajiv wrote:
> Dear Sir,
>     i am carrying outloadbalancing using OSCAR-3.0.We are also carrying out
> failover using HA-OSCAR 1.0 beta release (High Availability OSCAR). We are
> required to acheive loadbalancing
> and failover for the following services:
>     1. HTTP.
>     2. FTP.
>     3. TELNET.
>     4. DHCP.
>     5. SQUID.
> Our setup is as follows:
>   1 Primary server , 1 client node (using OSCAR-3.0)
>   1 standby server (using HA OSCAR)
> 
>     We have succeeded in building the cluster but am
> having problems regarding loadbalancing.We are trying to achieve
> loadbalancing using the PBS Server(Portable Batch System) which comes
> inbuilt with OSCAR-3.0.We are queing the services as jobs and trying to
> distribute these jobs between the server and  client node. But the problem
> we are facing is that we are not able to submit the job to the PBS server.
>     Sir,firstly, we would like you to confirm if we are going on the right
> track for achieving loadbalancing.We would like to know how you'll have
> achieved load balancing?
> 
> Regards,
> Rajiv
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 

-- 
Michael Will, Linux Sales Engineer
Tel:  415-954-2822  Toll Free:  888-PENGUIN   
Fax:  415-954-2899  www.penguincomputing.com
Visit us at FOSE 2005!
Washington Convention Center, Washington, DC
April 5th-7th, 2005
Linux Pavilion, Booth 2225




More information about the Beowulf mailing list