[Bioclusters] Re: Bioclusters Digest, Vol 11, Issue 11

Ryan Evans revans at hgen.pitt.edu
Thu Sep 29 12:18:50 EDT 2005


There is a file (/etc/systemserialnumbers/xsrv) that is just a plain  
text file containing the serial number for each server.
A post install script that runs on first boot could parse a list of  
serial numbers then echo the appropriate one into this file.
This would solve the issue of having to touch each machine to correct  
the problem.
(One would hope)

The best tool I found so far for imaging is here.
http://www.bombich.com/software/netrestore.html
It's kind of a cross between net-install and net-restore by allowing  
for the installation of a pre-configured os.


>
> Message: 2
> Date: Wed, 28 Sep 2005 15:15:22 -0400
> From: "Bergman, Andrew L." <andy_bergman at hms.harvard.edu>
> Subject: [Bioclusters] Imaging Xserves as compute nodes: OSX Server?
> To: <Bioclusters at bioinformatics.org>
> Message-ID:
>      
> <A35FED661C88CC40B7328CCFFCA6989F03C378F8 at MAILSERVER01.MED.HARVARD.EDU 
> >
>
> Content-Type: text/plain; charset="us-ascii"
>
> Hi,
>
> We're tryng to set up a straightforward way to image Xserves in an LSF
> cluster. I've had some success using NetRestore for basic brute-force
> imaging, and am also looking at radmind for maintaining incremental
> updates to systems. Right now we have only about 10 Xserves, but this
> number could grow.
>
> However, one thing I'm not clear on:
>
> OSX Server seems to require registration of the system to the OS  
> (by MAC
> address), at least for the Server functions to run, but possibly for
> other reasons, too. When imaging lots of systems, you generally use  
> one
> image (or a minimal number, anyway) to mass-deploy and try to make the
> procedure as automatic as possible, but is it necessary to touch each
> system for registration purposes for each installation/upgrade? I  
> really
> hope not, but I'm not sure what the recommended way around this is.
>
> My understanding is that Xserves require the Server version of OSX for
> the hardware to all work properly with the OS, though please  
> correct me
> if I'm wrong here. Only one of our Xserves needs to run any actual OSX
> Server processes (NetBoot, NFS) at this point, but I'm trying to avoid
> larger issues with doing things this way (i.e. conflicts from cloning
> "Server" on the same subnet). So, running OSX Client would probably  
> suit
> our needs for the compute nodes, but I'm not sure if that's viable?
>
> I'm posting to this list since there's folks with large Mac  
> clusters who
> I guess have figured out good ways to handle imaging. The Apple  
> docs are
> useful in many ways, but can be vague, too.
>
> Any suggestions for the best general tools for imaging the systems?
> Anyone using radmind?
>
> thanks!
> --andy
>
> ----------------------------------------------------------
> Andy Bergman
> Research Systems Administrator
> Research Information Technology Group
> Harvard Medical School
> phone: 617-432-7700
> fax: 617-432-5962
> http://ritg.med.harvard.edu/
> ----------------------------------------------------------
>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: http://bioinformatics.org/pipermail/bioclusters/attachments/ 
> 20050928/2a06a5cd/attachment-0001.html
>
> ------------------------------
>
> Message: 3
> Date: Thu, 29 Sep 2005 10:46:46 -0400
> From: Chris Dwan <cdwan at bioteam.net>
> Subject: Re: [Bioclusters] Imaging Xserves as compute nodes: OSX
>     Server?
> To: "Clustering,    compute farming & distributed computing in life
>     science informatics"    <bioclusters at bioinformatics.org>
> Message-ID: <4EF3B855-77C2-4805-94C7-28C602B34541 at bioteam.net>
> Content-Type: text/plain;    charset=US-ASCII;    delsp=yes;     
> format=flowed
>
>
> Apple has a licensing mechanism for exactly this use case, though
> it's not terribly well publicized even within their organization.  It
> it possible to obtain what they call a "watermarked" serial number
> for use within clusters.  This is a special OS X Server license which
> will not shut down server functionality if it is duplicated within a
> subnet.
>
> People can contact me off-list for the proper contact info to obtain
> one of these, if your local Apple rep hasn't heard of it yet.
>
> As to imaging, I've had luck with the System Image utility, and
> NetInstall (via netboot) in 10.4.  It's a lot better than any of the
> previous versions.  There is still an annoying behavior where a net-
> install will only re-image a particular *named* partition or disk.
> It's not possible (so far as I know) to automatically re-image the
> first disk device that they system finds.  This can necessitate
> booting the machines in firewire 'target' mode to rename partitions,
> which is pretty lame.
>
> I haven't tried running OS X client on an xserve.  I've stuck with
> server, mostly for access to the command line server tools for
> network settings and such.
>
> -Chris Dwan
>
> On Sep 28, 2005, at 3:15 PM, Bergman, Andrew L. wrote:
>
>
>> Hi,
>>
>> We're tryng to set up a straightforward way to image Xserves in an  
>> LSF
>> cluster. I've had some success using NetRestore for basic brute-force
>> imaging, and am also looking at radmind for maintaining incremental
>> updates to systems. Right now we have only about 10 Xserves, but this
>> number could grow.
>>
>> However, one thing I'm not clear on:
>>
>> OSX Server seems to require registration of the system to the OS
>> (by MAC
>> address), at least for the Server functions to run, but possibly for
>> other reasons, too. When imaging lots of systems, you generally use
>> one
>> image (or a minimal number, anyway) to mass-deploy and try to make  
>> the
>> procedure as automatic as possible, but is it necessary to touch each
>> system for registration purposes for each installation/upgrade? I
>> really
>> hope not, but I'm not sure what the recommended way around this is.
>>
>> My understanding is that Xserves require the Server version of OSX  
>> for
>> the hardware to all work properly with the OS, though please
>> correct me
>> if I'm wrong here. Only one of our Xserves needs to run any actual  
>> OSX
>> Server processes (NetBoot, NFS) at this point, but I'm trying to  
>> avoid
>> larger issues with doing things this way (i.e. conflicts from cloning
>> "Server" on the same subnet). So, running OSX Client would probably
>> suit
>> our needs for the compute nodes, but I'm not sure if that's viable?
>>
>> I'm posting to this list since there's folks with large Mac
>> clusters who
>> I guess have figured out good ways to handle imaging. The Apple
>> docs are
>> useful in many ways, but can be vague, too.
>>
>> Any suggestions for the best general tools for imaging the systems?
>> Anyone using radmind?
>>
>> thanks!
>> --andy
>>
>
>
> ------------------------------
>
> _______________________________________________
> Bioclusters maillist  -  Bioclusters at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters
>
>
> End of Bioclusters Digest, Vol 11, Issue 11
> *******************************************
>

Ryan Evans
Systems Programmer
Human Genetics Systems Support
Department of Human Genetics
University of Pittsburgh

There are 10 types of people in the world:
Those who understand binary and those who don't.

"Contrary to popular belief, UNIX is user friendly.
It just happens to be selective about who it makes friends with."


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://bioinformatics.org/pipermail/bioclusters/attachments/20050929/431ace19/attachment-0001.html


More information about the Bioclusters mailing list