Using LCFGng (LCFG new generation) it is possible to install and configure automatically a set of machines (a farm) from scratch. More important, the system allows to manage the configuration that you have and to recreate the exact settings.
This tool is useful not only if you have a large number of nodes but also if you want to automate your testbed installation and configuration. It is also possible to perform automatic upgrading/downgrading operations.
LCFGng configuration templates and tools (LCFGng components) have been developed in order to install and configure all the types of testbed machines: Computing Element (CE), Worker Node (WN), Storage Element (SE) and User Interface(UI) (for the Resource Broker and the Network Monitoring machine RPM lists exist but a manual configuration will be needed until LCFGng is introduced).
LCFGng needs a central server (LCFGng server) from which LCFGng clients fetch configuration files and RPMs (via HTTP and a NFS export).
On the LCFGng server both HTTP and DHCP servers are needed, and a disk partition is shared with the clients via NFS. The clients, during the boot process, take the IP number via DHCP and the configuration via HTTP.
The information about the server and clients installation is too long to fit in this document. The following sections are a short summary of the main points of the LCFGng server and client installation. Details can be found in the referenced documentation.
Soon the current 1.4x version of EDG will be ported to LCFGng. Information about the changes can be found on the WP4 website.
To install testbed machines via LCFGng the first step is the creation of an LCFGng server.
The WP4 web pages provide all of the information necessary to install a LCFGng server.
The server contains the operative systems for the clients and all the DataGrid RPMs to be installed in the clients. It also holds the client configuration files derived from templates.
The Notes about farm installation by means of LCFGng contains information about LCFGng installation of testbed machines.
The "edg-release" module, available via CVS, provides RPM lists and the configuration files.
In this repository you can find the RPM lists in the "ng_rpmlist" directory and the configuration files in the "ng_source" dir. A few of these files must be customized following your site-specific configuration. In particular the file "site-cfg.h" contains almost all the site-specific configurations, including the Globus configuration parameters (there is an example in CVS). The LCFGng components automatically configure Globus and DataGrid middleware reading the parameters in this file.
The meaning of the different Globus and DataGrid middleware parameters for CE,SE,WN and UI are explained in subsequent sections of this guide.
After the installation LCFGng also sets up the NFS home directories sharing between the CE and the WN.
Two items are not yet fully automated via LCFGng. The system administrator must manually:
Depending on the type of node, ranging from few to many additional manual configuration steps are involved.
Node | Function | RPM List | Manual Configuration |
CE | Gateway and local job control | yes | only local batch system |
WN | Worker node | yes | no |
UI | User interface | yes | no |
NM | Network monitoring node | yes | yes |
RC | Replica Catalog | yes | yes |
SE | Storage Interface Element | yes | no |
BDII | Information index node | yes | yes |
MP | MyProxy server | yes | no |
RB | Resource broker | yes | yes |
MDS | Top node of the MDS hierarchy | yes | no |