Skip to main content

Globus Transfer

February 11, 2019

With the increase in Globus usage, we have noticed a number of issues slowing or stopping transfer progress. While Globus is great at transfer persistence through errors, and is typically a "set it and forget it" utility, it is generally a good idea to keep an eye on your progress using Globus Activity Monitor.

Recommendations for successful transfers are: 

  1. Check Permissions - This is especially true in /projects directories where group collaborators may have used previously recommended transfer methods (e.g. rsync) which will have set incorrect permissions. Currently, permission errors will halt a transfer task without notice. Please be sure to cancel and resubmit transfer tasks with permission failures.
  2. Credential Lifetime - The default credential lifetime is 24 hours. This is fine for most small and moderate transfers, but may not quite be enough for large and/or delayed transfer tasks. When activating endpoints, please use the "Advanced" option to specify lifetime in hours. Maximum lifetime is 168 hours (7 days.)
  3. "Endpoint too busy" error - With the large amount of data moving from Peregrine to Eagle to MSS, the endpoints can be overwhelmed. We anticipated this and setup 3 Eagle endpoints. If you notice you are receiving an "endpoint too busy" error, please cancel your transfer task and restart using a different nrel#eglobus[1-3] endpoint.

Please see the NREL HPC documentation for Globus here for more information. You may also find the Transfer and Sharing FAQ helpful.