Requesting Assistance

From UMass GHPCC User Wiki
Jump to: navigation, search

Contact information

Please contact hpcc-support@umassmed.edu for questions or requesting assistance with GHPCC.

Hours of operation

MGHPCC staff is available from 8:30AM to 5PM on week days except for UMass-designated holidays. You will receive a reply to your request within 8 business hours.

The first Wednesday of every month is reserved for routine maintenance (all day, 9AM to 5PM). In the event there's maintenance that will extend past this window or has to be done at a different time, we will notify the user list 30 calendar days ahead of time so all users can take appropriate action.

Even with modern and mostly reliable systems, jobs can die for various reasons. If you expect your job to run for a very long time (3-4 weeks) you should see if you can either break the job up into smaller parts or see if it supports checkpoints. In the event your job dies you can restart from the checkpoint rather than lose all of the time you waited.

Items you can request support for

Support from MGHPCC technical staff is limited to the following:

  1. Queueing system (e.g. unable to submit jobs to the cluster, a job unexpectedly terminates)
  2. Authentication (e.g. unable to log into MGHPCC cluster, requesting new accounts, group membership)
  3. MGHPCC storage (e.g. quotas, assistance moving data on and off the cluster, permissions)
  4. Compute node problems
  5. Software (installing new software versions or libraries)
  6. Technical information about MGHPCC for grant applications

Items you will need local support for

Given the diverse community that makes up MGHPCC, we are unable to provide assistance with the following:

  1. Detailed application assistance (e.g. using BLAST)
  2. Local networking issues (please consult with your local IT support first if you are unable to reach MGHPCC)
  3. Linux programming or training

Requesting help from MGHPCC

When you fill out your request, please consider the following:

  1. Be as specific as possible. If a compute node is failing or a job is having a problem, include the job ID or compute node in question. Dates when you started noting the problem gives us more information.
  2. We do not need to know your password, but please include your username and any colleagues who are seeing the same issue
  3. If there is a way that we can easily replicate the problem, please include it. The quicker we can experience the problem ourselves, the faster we can determine the problem and ensure it's fixed
  4. Similarly, if you ask us to install software, please include a method for us to test and ensure it's been installed properly.
  5. Please respond to our questions quickly. We understand you're busy, but until we have a full understanding of the problem there isn't much we can do. We are making a great effort to respond to you quickly as well.