I’m currently trying to run a multi-copy peptide in solution simulation in SIRAH Force Field, and during the solvent equilibration stage, I’m getting hit with segementation fault error.
starting mdrun ‘Protein in water’
100000 steps, 2000.0 ps.
[hpc3-20-00:170256:0:170262] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffff97d8430)
==== backtrace (tid: 170262) ====
0 0x0000000000012cf0 __funlockfile() :0
1 0x0000000000df28d9 spread_on_grid() ???:0
2 0x000000000001da56 gomp_thread_start() ???:0
3 0x00000000000081ca start_thread() ???:0
4 0x0000000000039e73 __GI___clone() :0
Wondering if anyone else had the same issue? I tried older versions but the revived the same error. I’m mostly following a tutorial here: GROMACS — SIRAH Force Field documentation
Yes, the tutorial file works. This issue appeared when I try to run the simulation with multiple copies of the same peptide/protein
s. Also in the past when I ran a similar simulation of small molecules in SIRAH, I got around this issue by downloading the file from the cluster to a Mac and ran this step there before bringing it back to the cluster.
We try to avoid Segmentation faults in GROMACS, but it’s still difficult trying to avoid all of them (without performance penalties as well). Since it works with the provided tutorial systems, I would suspect that the cause of your problem is related to your input coordinates, i.e., that it’s a variant of a system explosion. Have you checked that the coordinates look fine, also when displaying periodic images of the system?
I don’t know of any. If there is no apparent overlap, or bonds that get distorted across the periodic boundary, then that’s at least a good sign.
You say that you get this crash during equilibration. I presume that the energy minimization worked fine, right? Was there any indication of large energies or forces?
Yes, the energy minimization works fine, and it looks like once I hit a certain number of copy of the hexapeptide, I get a segmentation fault regardless how spaced out I tried to place them. For example, 115 peptides in a 13x13x13 box works fine, but 116 peptides doesn’t work regardless of box size. I haven’t paid too much attention to the energy, but I’m still testing and will look into the energy more.
I don’t understand why the number of molecules would matter. How do you insert the molecules in the box? Do you ever resize the box after inserting molecules?
Regarding the energy minimization, it’s good to keep an eye on if the energy minimization converges, but also if there are large forces in the system. If a system is crashing during equilibration, it’s often worth having a look at (in a visualization program) the surroundings and bonds of the atom(s) with the highest force by the end of the EM.