CHECKLIST FOR INSTALLING MCFARM

This document describes the overall procedure for using the instructions for creating a D0 production farm optimally, i.e. hopefully with the minimum amount of work!

Overview of the Instructions:

There are seven documents describe how to install and configure the mcfarm package on your farm and configure it for MC production, not including the initial planning document that was sent out earlier. These include instructions on how to:

Ø Install Fermi LINUX RH 7.1

Ø Configure specific LINUX services for mcfarm operation

Ø Create an mcfarm job server.

Ø Create a generic mcfarm production node, which is where the D0 binaries are actually run, using an automated install script that we provide.

Ø Create a file server for your farm out of an existing production node.

Ø Create a gather server for the farm out of an existing production node. This will be used to bring an additional machine to bear on the task of gathering output to SAM at FermiLab.

The farm will have one job server, 2-3 file servers typically, and 1-2 gather servers apart from the job server. The rest of the machines in the farm will be production nodes, and most of the time spent during construction of the farm will be in creating the production nodes. We have therefore provided an automatic install script that will configure a given node as a production node once a job server has been setup.

We strongly recommend the following procedure in building the farm:

1. Resource Planning: It is highly recommended that you decide at the very outset on how the machines ear-marked for the farm are going to be used. Most importantly, you should identify which machine is going to be the job server, how many file servers you are going to have in the farm, and which machines are going to be the file servers. To make this decision refer to the document that was sent out initially on planning. The UTA team would be glad to help you in making this decision. Making the decision on the file servers and building them first greatly simplifies the task of building the farm. You don’t really have to decide on the gather server at the outset.

2. Once you have decided on the job server and the file servers, use the Linux Install document to bring up Linux on these machines.

3. Then use the Specific Services config document to setup some network services crucial to the farm as described.

4. Then configure the machine you have identified as the job server using the document provided.

5. Then you need to set up the file servers. The file servers initially need to be setup as production nodes and then converted to file servers. For each file server machine, use the instructions provided in the production node auto-prep document, and the automated script to configure it as a production node. Then use the file server-prep document to configure it as a file server.

6. Once you have all your file servers ready, then you can go ahead and start building the production nodes. To do this, install Linux, and configure the services required by mcfarm as in Steps 2 & 3. Then use the production node auto-prep document, and auto-config script to setup the production node. This greatly simplifies the task of building your farm as the auto-config script now has ALL the information it needs to bring this node up as a production node since the file servers are up as well as the job server.

7. Once all the production nodes are up, decide on the gather-server(s). Then use the gather-server preparation document to configure any production node(s) you want as a gather server.

After all this, you have a working D0 MC production farm and can start participating in the testing of the D0 Grid.

Good Luck!!

THE UTA TEAM