Re: upgrading the University of Richmond cluster

[gilfoyle <ggilfoyl@richmond.edu>]

hi steven,

   thanks for the quick response. our linux person at richmond is
starting
to work on steps 1-4 and thinks he may be able to get them done today.
we'll let you know as soon as he is done. i am at jefferson lab in 
newport news today, but i'll be back in richmond this evening and i'll
be around next week to work on the upgrade. if you want to talk to our
linux person directly his contact information in the following.

Sasko Stefanovski
804-287-6617  (office)
804-289-8784  (cluster lab)
sstefano@richmond.edu

thanks-in-advance,

jerry
 

steven james wrote:
> 
> Greetings,
> 
> For minimum disruption, I propose that the upgrade proceed in several
> steps.
> 
> 1. Secondary master shut down and disconnected from cluster (ethernet
> cable).
> 
> 2. re-configure BIOS to boot from CDROM, then hard drive
> 
> 3. Install RedHat 7.2 (later versions not advised at this time)
> 
> 4. Make sec. master available on public network for access by LinuxLabs.
> 
> 5. I will install Nimbus enhancements on secondary master and do
> preliminary configuration.
> 
> 6. back up all user data on master and compute nodes to fileserver.
> 
> 7. Primary master shut down. re-designated as secondary master
> 
> 8. Secondary master connected to cluster and re-designated as master.
> 
> 9. Final configuration. Some local intervention will be needed to boot
> compute nodes.
> 
> 10. new sec master set for PXE boot in BIOS and booted.
> 
> 11. I will re-configure sec master. Primary's image will be copied to
> secondary.
> 
> For step 3 (RedHat install), a special partitioning will be required.
> 
> use fdisk to partition hda! Partition as follows (order IS important!)
> hda3 500M
> hda2 500M
> hda1 (as desired for scratch, not less than 1G)
> hda4 extended, rest of drive
> 
> hda5 3G
> hda6 rest of drive
> 
> Note that 5 and 6 are extended partitions inside hda4.
> 
> The partitions should be created in the order listed here in order to make
> sure booting will work! Druid will tend to get it wrong.
> 
> Then, edit the partitions to specify filesystem and mountpoint as follows:
> 
> hda1 ext3fs /scratch
> hda2 swap
> hda3 ext3fs /
> hda5 ext3fs /var
> hda6 ext3fs /usr
> 
> Package installation:
> Select Custom install.
> Keep all default selected packages except for dialup
> 
> In addition, select:
> 
> NFS server
> development tools
> kernel development
> firewall/router
> 
> You may also select any other task/package sets desired.
> 
> In firewall setup, allow at least ssh (others as you see fit)
> eth2 MUST be set as a trusted interface.
> 
> Continue with installation as usual. Send me and Markus IP address and
> (temporary) root password when ready (step 4).
> 
> G'day,
> sjames
> 
> On Wed, 30 Oct 2002, gilfoyle wrote:
> 
> > Hi Steven,
> >
> >    What would be a good to arrange for an upgrade of the University
> > of Richmond cluster to the Nimbus system?
> >
> > cheers,
> >
> > jerry gilfoyle
> >
> >
> 
> --
> -------------------------steven james, director of research, linux labs
> ... ........ ..... ....                     230 peachtree st nw ste 701
> the original linux labs                             atlanta.ga.us 30303
>       -since 1995                              http://www.linuxlabs.com
>                                    office 404.577.7747 fax 404.577.7743
> -----------------------------------------------------------------------

-- 
Dr. Gerard P. Gilfoyle
Physics Department                e-mail: ggilfoyl@richmond.edu
University of Richmond, VA 23173  phone:  804-289-8255
USA                               fax:    804-289-8482