Setting a new machine (nmi-x.cs.wisc.edu) to be NWO ready

 

Configuring cfengine on a nmi-x system from scratch (As "root")

Execute the steps given below on nmi-x (steps preceded by grandcentral: need to be executed on grandcentral) in order.

Note: If Sean has already setup cfengine then jump to step - 3

  1. Install cfengine on the system. If it is a redhat system, Sean has already created the rpms which could be found at vger:/share/updates/cfengine. if it grumbles about the version of db software being installed, upgrade it to the required one and if that doesn't work, create a link (ln -s /lib/libdb-<major_version>.<minor_version>.so /lib/libdb<major_version>.so)
  2. Run cfkey to generate a public-private key pair.
  3. Check nmi-x:/etc/hosts and see if it has valid entries. (This is invariable wrong and set to public interface)
  4. Check if nmi-x:/var/cfengine/bin exists and has links to cfengine binaries.
  5. Verify if nmi-x has partitions as per the NWO standards
  6. Copy grandcentral:/cfengine/masterfiles/inputs/update.conf to nmi-x:/var/cfengine/inputs.
  7. If there are actions you need to take specific to this new system, modify grandcentral:/cfengine/masterfiles/inputs/cfagent.conf accordingly.
  8. On grandcentral add the hostname of this new system to grandcentral:/cfengine/masterfiles/inputs/cfrun.hosts and grandcentral:/var/cfengine/inputs/cfrun.hosts
  9. Edit/Verify grandcentral:/cfengine/masterfiles/inputs/cfagent.conf to add a class <arch>.<os>.<version/distro> under the section "Condor release file name". Look for examples in this section for other platforms. Put correct version of condor_release_bin, java_release_bin and any other release_bin that is compatible with this new platform. If you don't know or are not sure what cfengine class will distiguish this platform. run "cfagent -qv" and it will list the defined classes.
  10. Create file grandentral:/cfengine/masterfiles/configs/nmi-x.local and write the local condor configuration to it.
  11. If condor is run in the compatibility mode make sure that nmi-x:/prereq/condor-x.x.x has a valid condor installation. If it does not contain a valid installation stop condor and delete this directory.
  12. Run "cfagent -qv" and see if you get any errors. Correct them. If for some reason the installation of prereq failed just delete the /prereq/<software> directory before running "cfagent -qv" again
  13. Run cfservd on the machine.
  14. Once the condor daemons have started and the startds report to grandcentral, make sure the "uname" and "nmi_platform" hawkeye modules report correct info. If it is new platform then you may need to tweak the hawkeye modules. Put their updated copy in grandcentral:/cfengine/masterfiles/hawkeye_modules and also check them into cvs. Run cfrun to push the updates.
  15. Talk to Nick LeRoy and make sure your changes to uname module gets into the hawkeye distribution examples in condor.

Setting up the LDAP users database

Check with Sean for now.