log may seem like a lot to send, but the more information we have, the more likely we are to be able to diagnose and/or replicate your issue. Information that can help us includes:
- The cluster you're using
- The modules you have loaded
- The exact commands you're running
- Full error logs
- How you're connecting to the system (is it through a terminal, Open OnDemand shell, or virtual desktop?)
- Your batch submission script. This can be included either as a text file or by providing its full path and filename on the system.
Use reply-all to email chains - We will cc hpc-consult in our responses so that our (small) consulting team is able to view the issue and contribute.
Provide context - There's a common support issue called The XY Problem. Say you have a problem and try to solve it yourself but the attempted solution produces an additional problem. Submitting a ticket requesting help with the attempted solution without information about the original issue can lead to more confusion. Knowing the full context will help us be able to more effectively assist you.
Submit a ticket for your questions - If you reply to general system announcements or send emails to an HPC staff member's private inbox without cc'ing hpc-consult, your ticket may get lost in the shuffle and go unanswered (especially if your ticket goes to a staff member who has retired or is on vacation). Submitting a ticket will ensure we have a record of your question and will get to it as promptly as we are able.
What Are Our Support Policies
Mostly it is common sense rather than strict rules. The primary consideration is that our consultants work regular hours with some flexibility built-in. So don't expect detailed responses at night or on the weekends. You might get a quick response but don't count on it.
Our consultants typically don't know how to run your applications unless they have broad usage like Python or R. So once we determine it is likely an issue with the code we will refer you to the provid. We don't troubleshoot bugs or run profilers, but we support tools like Valgrind for you to use.
We don't mind you asking lots of questions. We encourage you to ask for a consulting session via zoom or Office Hours (see below). We are not in the office so you can't drop by (although we kind of miss that personal engagement).
The bottom line is that the supercomputers are only really productive tools when you have the support to gain the most out of them to improve both your results and the time to get results.