Revision 4 as of 2006-08-22 12:58:09

Clear message

Galaxy Administration

This page lists information relating to managing [http://galaxy.aero.iitb.ac.in/ Galaxy Cluster].

Services

Galaxy users following services. Some of the services are used for installation and setting up of compute nodes.

Service

Description

NIS

Galaxy users NIS/YP for sharing authentication information among frontend and compute nodes. Related services - portmap, ypserv, ypbind, yppasswdd. All maps are in /var/yp.

NFS

/home on galaxy is shared using NFS among all the compute nodes. Quotas are effective on compute nodes also. Related services - portmap, nfslock, nfs. Configuration in /etc/exports.

DHCP

All compute nodes get their IP address using DHCP. The database is statically configured. Related services - dhcpd. Configuration in /etc/dhcpd.conf.

EMAIL

Email is configured to forward root messages to amitay@aero. Basic sendmail configuration. Related services - sendmail. Configuration in /etc/mail.

HTTPD

Apache is configured to act as the web server. Web pages contain information about galaxy, status updates, news, etc. All web pages are located in /home/web. Most of the pages are PHP scripts. Related services - httpd. Configuration in /etc/httpd.

DNS

DNS is used to resolve all the compute node names. DNS database is in /var/named. Related services - named. Configuration file /etc/named.conf.

INETD

rsh, rlogin and time services are enabled through xinetd. Related services - xinetd, rsh, rlogin, time. Configuration files in /etc/xinetd.d.

TORQUE & MAUI

Torque is used as the batch jobs processing software and Maui scheduler is used. Torque server is configuring using qmgr command. Torque server (/usr/local/sbin/pbs_server) and Maui Scheduler (/usr/local/sbin/maui) are started from /etc/rc.d/rc.local. Torque configuration in /usr/spool/PBS. Maui configuration in /usr/spool/PBS/maui.

APC

APC PowerChute software is installed to shutdown galaxy cluster in the event of power failure. [http://ups-galaxy.aero.iitb.ac.in/ APC UPS] can last for up to 1 hour on full load when power fails. After which galaxy will shutdown all the compute nodes and frontend. Related services - PowerChute. Software installed in /usr/local/!PowerChute.

Managing Users

Galaxy Cluster can be used by any IIT personnel. All the applications for students have to be sent through the guide, for staff through the faculty in-charge of the project. Please collect [http://galaxy.aero.iitb.ac.in/?page=contact all the infomation as mentioned] on the galaxy website.

Following steps are required to add new user to galaxy.

useradd -c "Full Name" -d /home/<dept>/<username> <username>
passwd.orig <username>
cd /var/yp; make
edquota <username>       # block quota 1 GB soft, 1.1GB hard
vi /home/web/usage.php   # add username (email) 

Also add user to the [http://lists.iitb.ac.in/mailman/admin/galaxy-users/members/add galaxy-users mailing list].

Software Management

Galaxy is configured using Fedora Core 3 (FC3) Linux distribution. The installation of compute nodes is done using kickstart mechanism. (Insert the FC3 Install CD and type kickstart at the boot prompt.) Kickstart configuration file is /home/install/kickstart/ks.cfg. This file contains the instructions to setup compute node and install basic software required on each compute node. NFS Installation is done using /home/install/fedora directory that contains the entire FC3 distribution.

All the computational software installed on Galaxy is located in /usr/local (which is linked to /home/local) and is available on all compute nodes through NFS. Sources for all the installed software are kept in /home/install/download.