I have two very simple cases where structure predictor is still running after more than a week. They are binary metal chlorides! Could this be a bug?
Hmm, thanks for reporting. I see the still-running prediction jobs, and I see that you have several other jobs submitted more recently that have completed. We’ll look into this Monday morning.
P.S. We’re currently preparing for scheduled maintenance tomorrow of our infrastructure, so new submissions aren’t being pulled to run at this time.
What is happening is that the jobs are timing out. We do not actually run jobs until they finish; for structure prediction jobs, we allocate two hours of “wall time” when submitting the job to a queue on the supercomputing platform we use at NERSC. This is normally sufficient for structure prediction jobs.
In your case, the two cases timed out. We could certainly provide a better signal of this status, because the jobs are technically no longer “running”. Although a job may simply be restarted with a higher time allocation with the status still “running” for the end user, we do not in fact automatically restart structure prediction jobs. Sorry about this confusion.
We tried rerunning the jobs, and again hit the same wall time limit. Perhaps these jobs take longer because of their simplicity, i.e. there are more substitutions possible, but we don’t know yet. We will try re-running with a higher time limit and troubleshooting further if needed. Thank you for your patience.
After re-running with a higher time limit, one of the jobs failed and one again timed out. There is something peculiar about these jobs that we do not yet understand, and we will continue to troubleshoot. We see you have successfully run a handful of prediction jobs in the meantime. We’re glad you find the service useful, and we’ll let you know when we have a handle on why these particular jobs are not completing.
tried again today - - this one is still not working. seems that it does not work for binary metal chlorides