1. Donald Petravick

Donald Petravick

Forum Replies Created

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • Donald Petravick
    Participant

      We’ve been able to work together by sharing the slice ID,

      in reply to: Potential fix on certain bash command in notebook #3402
      Donald Petravick
      Participant

        @ilya @paul, so I moved to NERSC where Greg Daues is working successfully (and where we’d work to tie our prototype to the larger system,.) The good news is it works! (well almost). If Greg and l land on the same cluster head node, we conflict in /temp/fabric. (one of us owns the directory, the other cannot access it) I put a two line change fablib.py in my distribution to be sensitive to an environment variable to keep working. (see below with some surrounding lines for th change) I see there is a framework that tries to get FABRIC_LOG_FILE from a configuration setup, But FABRIC_LOG_FILE is not set-able from the environment. AFAICT.
        is there a supportable way or me to change the logfile/data directory in the current code? If not are my changes useful to you?

        all in all this is working out very well for us.

        LOG_LEVELS = {
        ‘DEBUG’: logging.DEBUG,
        ‘INFO’: logging.INFO,
        ‘WARNING’: logging.WARNING,
        ‘ERROR’: logging.ERROR,
        ‘CRITICAL’: logging.CRITICAL
        }

        default_fabric_rc = os.environ[‘HOME’] + ‘/work/fabric_config/fabric_rc’
        default_log_level = ‘DEBUG’
        default_data_dir = os.environ.get(‘FABRIC_DATA_DIR’,’/tmp/fablib’) #change near line526 fablib.py
        default_log_file = os.path.join(default_data_dir,’fablib.log’) # change

        fablib_object = None

        ssh_thread_pool_executor = None

        in reply to: Potential fix on certain bash command in notebook #3354
        Donald Petravick
        Participant

          no, the test is

          two nodes at NCSA,

          two at TACC. Each with

          a L3 IPV4 network

          after the submit, I go to set up routes so that all nodes talk to all other nodes.

          When I get the network object to find things I need for routing the network.get_gateway() and network.get_subnet() calls  give the errors like

          (Pdb) network.get_subnet()
          WARNING:root:Failed to get layer: ‘str’ object has no attribute ‘layer’

          It broke on Friday, so I suspected a change in the downtime,  but after Greg and I worked together today, I see the software still works at NERSC, so I am beginning to wonder about my (rickety) python install on my mac…  So let me clean house a bit and see what I learn.

          in reply to: Potential fix on certain bash command in notebook #3350
          Donald Petravick
          Participant

            Similarly,  I am working with the libraries on Linix/MAC.    I had working  code break as of Friday.

            the submit emits errors:

            WARNING:root:Failed to get layer: ‘str’ object has no attribute ‘layer

            My attempt to  set a route after submit

            (Pdb) network.get_subnet()
            WARNING:root:Failed to get layer: ‘str’ object has no attribute ‘layer’

            I’m no expert, but the network.slice.slice is serialized java and looks like all the network provisioning was done on the server,  but maybe on Friday, there was a change on the FABRIC server-side on how all this was serialized when it was sent to the client, and not it’s not de-serializing?

            Donald Petravick
            Participant

              hi our issue is simpler.  we don/t  need IPv6 right now, we are just looking to see if it is a bug that we cannot use these sites, our lack of knowledge on how to configure these sites  or should we just use TACC. at the moment just trying to get DNS to resolve the names.

            Viewing 5 posts - 1 through 5 (of 5 total)