Difference between revisions of "Administrator's Guide: How to set up a gCube infrastructure"

Revision as of 01:44, 2 October 2009

A gCube infrastructure is a set of working nodes (so-called gHNs, gCube Hosting Nodes) glued by the gCube enabling services and able to host gCube services in a cooperative way. When creating a new infrastructure, there are two kinds of configuration: secure configuration and non-secure configuration.

Non-secure configuration

Minimal deployment scenario

In this section, we report the steps to setup a gCube infrastructure with one Virtual Organization (VO):

1. decide the scope hierarchy configuration, i.e. the names of the Infrastructure and of the VO
2. identify a set of machines to turn on as gHNs (their number may vary depending on the infrastructure needs)
3. prepare a Service Map file for each scope (Infrastructure and VO)
4. setup the Infrastructure:
   1. identify 3 machines to dedicate to the Infrastructure Enabling Services
   2. install gCore in the 3 machines and copy the two Service Map files under the $GLOBUS_LOCATION/config folder
   3. configure the gHN to join to the infrastructure scope and
      • deploy the IS-Collector on the node
   4. configure the second gHN as ROOT and to join to the infrastructure scope and
      • deploy the IS-Registry on the node
      • deploy the IS-Notifier on the node
   5. configure one third gHN to join it to the infrastructure scope and
      • deploy and populate a Software Repository there
      • deploy a VREManager Service there
   6. start the 3 containers following the order of the deployments and verify that they work properly
5. setup the VO
   1. identify 3 machines to dedicate to the VO Enabling Services
   2. install gCore in the 3 machines and copy the Service Map files under the $GLOBUS_LOCATION/config' folder
   3. configure one gHN as root, join it to the VO as startScope and install the IS core services (IS-IC, IS-Registry, IS-Notifier) there dedicated to the VO (named IS VO gHN)
   4. configure one gHN (named VREManager VO gHN) to join it to the VO as start scope and install and configure a DL Management Service there

1. 1. configure the gHN to join to the VO scope and
      • deploy the IS-Collector on the node
   2. configure the second gHN as ROOT and to join to the VO scope and
      • deploy the IS-Registry on the node
      • deploy the IS-Notifier on the node
   3. configure one third gHN to join it to the VO scope and
      • deploy and configure a VREModeler to join the VO scope
      • deploy and configure VREManager Service to join the VO scope
   4. start the 3 containers following the order of the deployments and verify that they work properly

1. 1. start the container on the IS VO gHN and then on the VREManager VO gHN and verify that they work properly
2. configure and start generic gHNs
   1. install gCore in each machine and copy the VO Map files under the $GLOBUS_LOCATION/config folder
   2. configure the gHNs to join the VO as start scope:
   3. start the container on each machine and verify that the gHN is correctly published both in the root VO and in the VO IS
3. setup the portal by following the ... instructions.
4. configure one or more VREs by exploiting the Administration user interface capability.

Other possible configurations

Alternative configurations can improve the infrastructure performances. In particular:

• the IS-Notifier can be hosted on a different (and standard, i.e. non-root) gHN with respect to the other two IS core services
• the Software Repository can be hosted on a different (and standard, i.e. non-root) gHN with respect to the VREManager service
• multiple VOs can be defined
• more than one IS can be setup for each VO in order to distribute the load over them

The 'optimal' configuration mainly depends on the number of available gHNs. More gHNs joining the infrastructure means a better distribution of resources and services across them.