sw3 switches then plot d3plots are stoped
leszpol
Dear All,
I am facing a serious problem. I am running an UNDEX simulation for a very large structure, and the computation takes around six days or even longer. I perform the simulation remotely on university Linux servers. Every 24 hours (or at any chosen interval), I create d3dump files in case I need to stop and later continue the computation.
To do this, I use the d3kil command with the sw3 switch, which correctly generates the d3full and d3dump files. However, after that point, the creation of d3plot files stops completely. If I use the sw4 switch, only a single d3plot file appears (triggered by the sw4 signal), and no further d3plots are generated afterward.
Could someone please explain how to resume d3plot output after using the sw3 switch?
I have read the manuals and asked in several places, but perhaps this time someone who knows the actual solution can help.
Regards,
Leszek
leszpol
I have still been studiing the problem further and would like to add a supplement to my previous post, as it may help others.
The issue concerns massively parallel computing on a Linux server using multiple nodes. It turns out that when the
job runs on a single node, everything works fine — d3plot files are generated correctly. However, when more than one node
is used, the problem appears and d3plot output stops.
Has anyone encountered the same issue and managed to solve it?
Leszek
l...@schwer.net
I think you are doing things correctly.
So maybe this is a version bug?
Perhaps try a more recent version of LS-DYNA. --len
--
You received this message because you are subscribed to the Google Groups "LS-DYNA2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ls-dyna2+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ls-dyna2/3ca31f54-8075-4311-8796-11a937c07d00n%40googlegroups.com.
leszpol
Thanks for your support. The cluster runs the newest version of LS-DYNA, and I have no influence over it due to governmental license rules. However, I think you are right — it must be a specific issue related to massive parallel computing.
I found some information indicating that this is a known problem, but so far there seems to be only one possible test option, which I’m going to try. If I find a solution, I’ll share the information.
leszpol
The sw3/sw4, d3dump/d3full, RUNRSF/RUNFULL problem is solved. Everything works. The apparent stall was due to my computation setup, not a cluster issue.
The documentation explains the procedure step by step. Follow it and it works. If any questions, please ask.
Leszek