These installation instructions apply to the binary distributions of version 5 Condor. These distributions are available in source form also, but we are not attempting to provide instructions on how to build condor from the source this time around. We have found that there is a great deal of variation of the availability and in the implementation of the tools we use to build Condor, and we can no longer keep up with the questions and problems that arise. We don't want to discourage anybody from building and modifying the code to meet their needs, it's just that if you choose to go this route, you're on your own. The binary distribution is packaged in the following 4 directories: bin - executable programs lib - libraries to link condor user programs and scripts doc - postscript and ascii versions of documents examples - C, Fortran, and C++ example programs You should plan to install "bin" and "lib" under a common directory in some well known location so that your users can find the condor programs and libraries in a consistent place. The doc directory can go wherever is customary at your site. The examples directory could also go anywhere, but the "Makefiles" do contain relative references to the "lib" directory. If you move the examples directories, you will need to change the "Makefiles" before you can build the examples. SOLARIS SPECIFIC NOTE: Condor no longer needs to run under the root user-id *on Solaris only*. Instead, all daemons may be run under the condor user-id. We do not recommend mixing root daemons with condor daemons. Some documents will still refer to the need for the root user-id. You can ignore those parts. 1. Un-compress and un-tar the distribution. Example: you want to install all the condor related stuff in "/usr/condor". 1. mkdir /usr/condor 2. cd /usr/condor 3. (fetch the distribution file) 4. uncompress condor_5.1a_DecAlpha.tar.Z 5. tar xf condor_5.1a_DecAlpha.tar 2. One machine in your condor pool will act as a "central manager" for the pool. You should decide which machine will serve that function, and install Condor there first. Because of the importance of this machine, we recommend you pick one which is likely to be reliable, or at least to get rebooted promptly if it does crash. 3. Each machine in your pool will need a "condor" account, and a "condor" group. Condor's UID and GID should be consistent across all machines in your pool. You can easily add machines to your pool at any time, but you should decide at this point where your condor home directories will be located. You should install the "condor" account for at least your central manager machine now. 4. Several of the programs must be "setgid condor". Go to the condor "bin" directory and make certain that the following programs have owner "condor", group "condor", and permission "-rwxrwsr-x": condor_globalq condor_history condor_jobqueue condor_preen condor_prio condor_q condor_rm condor_submit condor_summary condor_throttle.generic 5. If Condor's home directory will be shared across the machines in your pool, you will need to create a directory for each machine where it can keep machine specific data. Make the directories under Condor's home directory, and name each directory with the hostname of the machine whose data it will hold. If Condor's home directory will not be shared across the machines in your pool, you don't need to do this. 6. If condor runs into a problem at your site, it will send mail describing what went wrong. You need to decide who should get such mail. You may want to make this an alias so that you can change the recipient of the mail later without changing condor. 7. Once a week Condor will try to send a status report back to its authors. This function can be defeated, but if there is any way within reason that you can allow this at your site, we would really appreciate your cooperation. This kind of information is really helpful to us in determining (and demonstrating) the effectiveness of our work. Our internet mailing address is condor_admin@cs.wisc.edu please consider whether you can send mail to this address from your site, or whether you can make some modification of the address to get the mail delivered. You will be asked for this information later. 8. Some of the scripts and configuration files in the binary distribution will need to be customized for your installation. A program called "condor_customize" is provided for this (in the "bin" directory). The program will ask you 5 things, providing defaults for each. You should be prepared to answer these questions before you run the customization program. a. Which host you want act as the central manager for your Condor pool. See step 2 above. b. The local email address where you want Condor to notify you regarding problems. See step 5 above. c. The email address where condor can send a weekly status report back to its authors. See step 6 above. d. The pathname of the directory which contains the "bin" and "lib" directories. See step 1 above. e. The pathname of the directory which contains the machine specific data. See step 4 above. Note: two macros are available to simplify the specification of this directory. The $(TILDE) macro translates to the name of Condor's home directory on whatever machine it is evaluated on, and the $(HOSTNAME) macro evaluates to the hostname of whatever machine it is evaluated on. Thus if you have separate home directories for all the condor accounts in your pool you could specify $(TILDE) for this value, and if you have a shared home directory for condor you could specify $(TILDE)/$(HOSTNAME) 9. Build subdirectories to hold the machine specific data by running "condor_init" on each machine you want in your pool. N.B. you must be "root" when you run "condor_init" (except on Solaris). 10. Start the condor daemons by running "condor_on" on each machine you want in your pool. N.B. you must be "root" when you run this program (except on Solaris). 11. Ensure that condor is running. You can run ps -e | egrep condor_ on you central manager machine you should have processes for condor_master condor_collector condor_negotiator condor_kbdd condor_startd condor_schedd On all other machines in your pool you should have processes for condor_master condor_kbdd condor_startd condor_schedd (Suns and HP's don't run "condor_kbdd" as they dont' need it.) 12. Ensure that the condor daemons are communicating. You can run "condor_status" to get a one line summary of the status of each machine in your pool. 13. Try building and running some test jobs. A separate document describes how to link and run these jobs.