Groups.grid.org

Has anyone been successful on having the UniCluster authenticate against an external (public LAN) LDAP server? Hints and suggestions welcome. 

I am planning to install the UniCluster 4.1 on CentOS5.2. This is my first cluster install so please bear with me. According to the documentation it is recommended that the master node has two interfaces (public and private). My question pertains to using NAS for shared storage access (/home, etc.) I have a NAS attached to the public LAN. It is obvious that a NFS mount can be established between the master node and the NAS as it has a public LAN interface. Should the compute/workstation/login nodes also have two interfaces as well to gain access to the NAS?

To help rapidly deploy installer nodes during development we often use the standard Anaconda kickstart install mechanism.  This way we can, in an unattended manner, place hardware servers in a state that is ready to install unicluster.  Here is an example of one such kickstart file (this kickstart is hosted by a cobbler server that also manages the online repo hosting and mirroring of the CentOS update RPMS:

While the unicluster-setup script checks for several configuration errors on the Installer node it is better to check your configuration before attempting to run the setup script. Here is a list steps to help validate the configuration of your installer node:

http://continuations.com/post/61496327/kaizen-for-developers-no-inventor...

One thing I like about this article is that in Unicluster 4.1 we saw this coming.  You can have Unicluster get resources from EC2 whenever you max out your current Data Center.

If you installed the UniCluster Test drive on a machine and would like to remove it you can follow these step (all steps run on the installer node):

UniNework is a network software repository for UniCluster 4.1
located on servers at Univa UD. When you sign up at grid.org your
account is automatically associated with UniNetwork allowing you to
download UniCluster 4.1 and all of the free UniCluster 4.1 Pro-Packs
and software Kits.  If you purchase the optional UniCluster support
subscription your grid.org UniNetwork account will receive additional
network entitlements providing access to our UniCluster updates
repository. 

The Cluster Migration Pro-Pack is an optional pro-pack available to
UniCluster Support Subscription Customers.  The Cluster Migration
Pro-Pack contains several Documents and Tools that reduce the effort of
moving from LSF to SGE running in a UniCluster environment.

Cluster Migration Pro-Pack contains:

The Intel Cluster Ready Pro-Pack is an optional Pro-Pack for
UniNetwork customers building HPC clusters with Intel Processors.  The
Pro-Pack contains the Intel Cluster Checker a tool used to certify
Intel clusters under the Intel Cluster Ready Program http://www.intel.com/go/cluster.

HPC Clusters are certified as Intel Cluster Ready when the cluster is
deployed according to a Cluster Recipe then certified using the Intel
Cluster Checker.

Creating a new installer node

Hello All,
I decide to install unicluster express on a centos 5.2,I do step by step from installation guide.I have a problem, when installation is in processing this error apears:

Candidate must have Architect level experience with:

--Grid Computing concepts including Federated Grid Computing

--DataSynapse Gridserver

--Virtualization techniques including Zones

--LDAM (Logical Device Allocation Map)

--VMWare

--Linux/Unix

Location: Stamford CT
Duration: 6+ months
Rate: Open

Please contact:
Carl Noel
Sharp Decisions
212.481.5533x564
carl@sharpdecisions.com

Hi,
I was wondering how to go about contributing to, and observing the progress of developments in the UniCluster stack?
Specifically, there does not seem to be a way to obtain a latest copy of the UniCluster stack in development.

Would it be possible to place the code on github (github.com), where it is easy fork, and observe other's changes (the network diagram)? Or some other such facility where everyone can easily cherry-pick from anyone?

Mark

Moved from the EC2 forum/group (seemed dormant)

Hi,

I'm considering using Kepler (http://kepler-project.org) to develop workflows, and the Kepler project has a distributed workflow interest group (http://dev.kepler-project.org/developers/interest-groups/distributed/arc...)

Q1) Is Kepler, with or without its distributed actor, known to be working in the UCE stack?

Q2) Is such a distributed-actor use case a standard one for the UCE+SGE projects? Is this use case the subject of much effort, or is it considered to be a fringe usage, with not much work under way?

I moved this from the EC2 group/forum (seemed dormant):

Hi Group,
Thanks for the UCE+EC2 screencast and white paper - I've found them very interesting.

I'm relatively new to UCE, SGE and the Globus/grid infrastructure, though I have looked at them before, I have not used them.
I'm a non-computer-science domain an analyst, rather than sys admin or programmer. With that background in mind I have setup and used Condor, and seperately have 'toyed' with Amazon's EC2.

I have some questions I'd appreciate any comment/insight on:

Hi sorry to whinge.....

I'm really struggling to see the value added by having _both_ group and forum posts.... worse it seems to fracture the information and increase each individual's search costs. In addition there is a _ton_ of clutter.

It would be nice to have the ability to reply 'in-line' the a post in the forum.

Best, IMO, would be to use several Google groups named:
grid-

e.g. grid-newbies, grid-unicluster, grid-sge, grid-gurus, grid-ohmi, grid-ec2, grid-mpsub, etc

Hi Group,

In Amazon's EC2, you are limited to onje network interface and have a private IP and one public IP. The UC4.0 install instructions state:

"With a two network configuration one network interface is the public interface connected to the internet or intranet and the other network interface is used to create a private cluster network for all of the compute nodes. "

Would it be enough to create some alias, say eth0:1?

Mark

This is obvious to many of the SAs out there, but to the uninitiated a note somewhere would have helped...

Using a clean CentOS install, you must configure Yum to get through a Firewall (e.g., company Firewall).  Several files must be edited...

1) /etc/yum.conf, add the following line:

(if you work for yoyodyne inc., the following works.  Otherwise, ask your company network guys for your proxy host and port number)

http_proxy = http://proxy.yoyodyne.com:18023/

Please let us know if you have any issues with the installation or configuration of UnICluster 4.0 so we can work through the problems and improve the product.

Thanks,

Cameron

The UniCluster 4.0 product provides cluster provisioning functionality with the ability to extend its functionality with software "kits". For mor information on kits you can read the wiki article, http://wiki.grid.org/AdminGuide_Kits.

When evaluating cluster provisioning systems what features are most important to you?  Do you perfer imaged or diskless nodes?  Would you like the compute nodes to be as similiar to a stock distro install as possible?  How important is support for various distros?

 -Cameron

HI Friends,

I have been trying to install Unicluster Main node ( unicluster_express_3_2_0_0 ) in Opne Suse 10 ( 64bit OS). I have been facing one library missing issue during the installation. But I found that particular library is available.....

suse1:/opt/unicluster # sh install_unicluster.sh

The xinetd service is installed. Continuing installation.

**********************************************************************

Installing a local version of Perl to
/usr/local/UnivaUD/perl

This version of Perl is installed for UniCluster Express use only.

Years ago I subscribed to an email list that has been routinely filling up my personal email account on a daily basis (I honestly forget the source, or what I was thinking at the time). With nothing else to do at airports waiting for delayed planes (which seems to have become a regular occurrence) I started reading some of them (Have Blackberry will read!). A couple of days ago an email came in from internetnews.com that caught my attention: “What Bad Economy? IT Spend Will Grow This Year”.

from their web site: http://sc08.supercomputing.org/?pg=about.html