Version 1 (modified by, 10 years ago) (diff)


Appendix 1: Running a Private OML Server

1.1 OML

More detailed information on OML and the its installation can be found here.

1.2 Experiment controller

The experiment controller configuration file, located at /etc/omf-expctl-5.4/omf-expctl.yaml, specifies which oml server the nodes should communicate with. The snippet below show the current configuration that's in the file on the tutorial user workspace VM.

# URI to the OML server to use for this EC 
# (can be overwritten on the EC command line)
#:omluri: 'tcp:'
:omluri: ''

In case you were planning an experiment where the OML server is located somewhere else you would have to modify this line accordingly. You have to assure that the omluri is reachable from ALL nodes that participate in your experiment. This does not necessarily mean that the OML server has a public IP address. E.g., in the case of an experiment on an ExoGENI slice, the OML server can be located on one of the nodes of this slice.

1.3 OML server on ExoGENI node

The current ExoGENI image we currently use for the tutorial already comes with an OML server installed. This allows you to run an OML server of one of the nodes in your slice.

  • Run OML server on one of the nodes in your slice.
  • If you'd like to automatically store measurement results in iRODS you have to copy the hook script to /usr/share/oml2-server/ and start the OML server with the -H <path_to_hook_script> flag.
  • Configure :omluri: entry /etc/omf-expctl-5.4/omf-expctl.yaml on the node where you run the experiment controller. (Note: this can be on a machine that's outside your ExoGENI slice, as it is the case for this tutorial.)
  • Run your experiment.

IMPORTANT NOTE!!! Make sure that you safe your experiment in persistent storage (e.g., iRODS) before you take down your slice. If you only keep a version on our ExoGENI node it'll be lost after the slice is removed!! Back to Tutorial Section C