| layout | title | permalink |
|---|---|---|
default |
troubleshooting |
troubleshooting.html |
This page has information about troubleshooting MESA.
One of the most common issues is unset or incorrectly set environment variables. In the same terminal window where you are trying to run MESA, execute the command {% highlight bash %} echo $MESA_DIR {% endhighlight %} and if you're using the MESA SDK, execute the command {% highlight bash %} echo $MESASDK_ROOT {% endhighlight %} Confirm that these showed the directories where you have installed MESA and the MESA SDK. If they did not, please re-read the instructions on how to set your environment variables.
Check to see if there is any information about your problem in the MESA FAQ.
If you are using the MESA SDK and are having a problem with installation, you should also consult the MESA SDK FAQ.
Search the mailing list archives to see if someone has had a similar problem in the past.
Send an email message to [email protected] describing the problem. If you want a good answer, it helps to ask a good question.
Provide some information about the computer you are using, including:
- your operating system (including version information)
- which version of MESA you are using
- which version of the MESA SDK you are using
Then, describe the problem. Include sufficient materials (inlist files, saved models, etc.) so that someone else can (attempt to) reproduce your problem.
Avoid attaching screenshots of output or error messages. Rather copy the text into your message or, if there's a lot of text, attach it in a plain text file.
Perhaps someone else can give you some useful information -- or at least some sympathy.
It is an unfortunate truth that you often have to tune the inlist control parameters to fit the particular task you are attempting. How much time and effort that will require can vary from none when you are using a proven recipe from "off the shelf", to painfully long if you are being a pioneer and trying something that hasn't been done before.
Here are a few hints based on my experience.
Sometimes a failure to converge is simply caused by "grid starvation". Try giving the code more points to work with by decreasing the value of "mesh_delta_coeff" and restarting from a step before the convergence problems started.
Use your favorite visualization tools to get an idea of what's going on -- make lots of plots! I strongly recommend using PGstar or something like it to make on-screen plots as the run progresses.
If you can't find a recipe that works for your problem, try a different problem.
Okay -- enough doom and gloom. Sometimes things work great, and you'd like to see if you can make them even better by improving the speed. You will of course need to check that the important results aren't impacted.
Finally, remember what Peter Eggleton told me when I was first getting started and asked him why something I was trying failed to converge. Peter patiently explained that with stellar evolution the only surprise is when the code does converge! And of course he's right. Each step requires a root find for a highly non-linear relation involving anywhere from a few thousand to several 100's of thousands of variables. It makes me tired just to think about it.
And remember that it takes time to learn how to use the code -- you'll get better at it with experience.
-- Bill Paxton