Re: a FAQ(?) which I do not see covered in the FAQ...

Hi,

I have used openMosix and various failover mechanisms together but
have tried nothing as of yet to checkpoint migrated processes.

(I have had no need for checkpointing oM processes - as long as a node
doesnt die when critical processes have migrated, then everything is
fine - but if they do, I just run the process again)

I have 2 master nodes running heartbeat, DRBD, ipvs and ldirectord.
All nodes have the same openmosix kernel.
All nodes share the same /home directory
Worker nodes share the same nfs root

DRBD (with heartbeat) as a failover mechanism is great as with careful
tweaking, you can have a shared nfs root on worker nodes which will
failover if the nfsd node dies.

Using heartbeat to control failover situations makes for easy
configuration and can enable services to failover (eg. a master cron
daemon for the network controlling nodes using DSH to run commands)

heartbeat along side ipvs (ldirectord) is also a very good way to load
balance services and the master node will forward traffic to to the
worker nodes with weighted round robin/weighted least connections and
the like. (ie. I run apache on all my worker nodes and heartbeat with
ipvs on the master will use each one in turn based on number of
connections (wlc) so a number of apache servers are seen as 1 from the
outside world)

Also, ipvs will also failover to the master node if all worker nodes die.

I have recently been playing around with DRBD to allow ALL nodes to
share the same root with failover across the 2 master nodes (hence my
reason for straying away from 2.6oM kernel testing)

I would like to learn more about gfs (but thats a different story)

Anyways, hope that gives a few ppl some ideas on how to use oM along
side high availability.

Martin


On 21/11/06, Evert <[email protected]> wrote:
> Lars O. Grobe wrote:
> > P.S.: You might still use openMosix in a failover setup, if you bind the
> > critical services to the home node (and if that fails, they get started
> > on the failover system) and still keep running less critical processes
> > openMosix-controlled. That way you could use all processing power and
> > still have a controlled failover. Somewhat an exotic setup still...
> >
>
> Is there anyone out there who has such a setup successfully in use?  8-)
>
>
> Regards,
>   Evert
>
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share your
> opinions on IT & business topics through brief surveys - and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> openMosix-general mailing list
> openMosix-general-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> https://lists.sourceforge.net/lists/listinfo/openmosix-general
>

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV