Cloud howto: running batch jobs with HTCondor/Cloud Scheduler

• Obtain images - three methods:
  1. Use cloud provided default images (installed as public images by sysadmins) - easiest method.
  2. Download an existing image (see OpenStack image guide, section 2):
     • Select and image and copy its' URL
     • Use the horizon "Create Image" dialog and copied URL to upload the required image to your cloud account.
  3. Create image with virt-install (and cobbler):
     1. Hardware confguration
     2. On hypervisor node:
        • create and run /root/qemu/mk-cc-demo. This will create a virtual machine with a vnc console that will fail to boot. At this point, the vnc console can be closed.
        • Issue "virsh edit cc-demo" and copy the "MAC" address.
        • Modify /root/qemu/mk-cc-demo to include the copied MAC address.
        • "virsh destroy cc-demo"
        • "virsh undefine cc-demo"
     3. On cobbler server:
        • Create a public/private ssh key: ssh-keygen -f /root/.ssh/id_rsa_cc-demo
        • Create a /var/lib/cobbler/kickstarts/cc-demo.cfg suitable to your application needs.
          • Create an encrypted root password (make sure cipher matches passalgo) and insert in kickstart: mkpasswd -m sha-512 following [[][these instructions]].
          • Copy your new public key to the end of your kickstart file.
        • Create /root/cobbler/add-cc-demo specifying the kickstart file and the copied MAC address. Run the shell script and then "cobbler sync"
     4. On the hypervisor node:
        • Re-run the "mk-cc-demo" shell script. This will recreate the VM with a vnc console, but this time it will boot/install. * ( 5 minutes) *
        • After the reboot, prepare the VM for other network connections:
          • ssh login from the cobbler server: ssh -i ~/.ssh/id_rsa_cc-demo 192.168.1.11
          • /etc/sysconfig/network (no host or gateway, nozeroconf=yes)
          • /etc/sysconfig/network-scripts/ifcfg-eth0
          • /etc/inittab
          • halt
          • virsh edit cc-demo and change the network bridge to br3 (public network)
          • virsh start cc-demo (virsh start cc-demo; virt-viewer cc-demo)
        • Which public IP did it get?
          • reboot in single user mode
          • ifup eth0 and ifconfig
          • reboot
        • Log in through ssh and make any required customizations (see, OpenStack image guide, section 5):
          • yum -y update * (10 minutes) *
          • yum -y install epel-release
          • yum -y install cloud-init (may complain about not having python-six)
          • Batch jobs? Need to install HTCtcondor:
            • (cd /etc/yum.repos.d; wget http://research.cs.wisc.edu/htcondor/yum/repo.d/htcondor-stable-rhel6.repo)
            • yum -y install condor
            • Modify the daemon list in /etc/condor/condor_config to include only the MASTER and STARTD daemons.
          • Authenticating condor server/workers with GSI?:
            • Create /etc/yum.repos.d/igtf.repo and install the following package:
            • ca_policy_igtf-classic
          • rm -fv /etc/udev/rules.d/70-persistent-net.rules
        • shutdown the VM
        • Use the horizon "Create Image" dialog to upload the image to your cloud account.

• Customizing images interactively on the cloud:
  • Interractive instantiation:
    • Images->Actions->Launch
    • Provide name, flavor, key pair, network - then Launch
    • Instances->actions->associate floating IP
  • Modifying an image:
    • Login as root
    • Make changes
    • Snapshot:
      • Instances->Actions->Snapshot, provide snapshot name.

• Running batch jobs:
  • On CC (litai08) demonstrate the necessity for grid credentials:
    • condor_q
    • ~/.globus
    • grid-proxy-ini
    • condor_q
  • Submit jobs:
    • Prepare batch control files:
    • cc-demo.sh - executable
    • cc-demo.job - Job Description Language (JDL) file
    • condor_submit
  • Monitoring jobs:
    • bin/watch: cloud_status -m, condor_status -m, condor_q
  • Retrieving job output & removing jobs:
    • condor_transfer
    • condor_q & condor_rm
    • check our spool directory
  • Custom VM contextualization:
    • yaml/custom.yaml
    • scp yaml/custom.yaml unix.uvic.ca:www/
    • +VMAMIGONFIG.
    • condor_submit
    • condor_release
    • The job log.